Journal of Energy and Power Engineering

Journal of Energy and Power Engineering Volume 4, Number 8, August 2010 (Serial Number 33)

David Publishing

David Publishing Company www.davidpublishing.com

Publication Information: Journal of Energy and Power Engineering (ISSN1934-8975) is published monthly in hard copy and online by David Publishing Company located at 1840 Industrial Drive, Suite 160, Libertyville, Illinois 60048, USA. Aims and Scope: Journal of Energy and Power Engineering, a monthly professional academic journal, covers all sorts of researches on Thermal Science, Fluid Mechanics, Energy and Environment, Power System and Automation, Power Electronic, High Voltage and Pulse Power, Sustainable Energy as well as other energy issues. Editorial Board Members: Prof. Ramesh K. Agarwal (USA), Prof. Hussain H. Al-Kayiem (Malaysia), Prof. Zohrab Melikyan (Armenia), Prof. Pinakeswar Mahanta (India), Prof. Carlos J. Renedo Estébane (Spain), Prof. Mohamed Ahmed Hassan El-Sayed (Trinidad and Tobago), Prof. Carlos Redondo Gil (Spain), Prof. Roberto Cesar Betini (Brazil), Prof. Rosário Calado (Portugal), Prof. Dr. Ali Hamzeh (Germany). Manuscripts and correspondence are invited for publication. You can submit your papers via Web Submission, or E-mail to [email protected] or [email protected]. Submission guidelines and Web Submission system are available at http://www.davidpublishing.com. Editorial Office: 1840 Industrial Drive, Suite 160 Libertyville, Illinois 60048 Tel: 1-847-281-9826 Fax: 1-847-281-9855 E-mail: [email protected]; [email protected] Copyright©2010 by David Publishing Company and individual contributors. All rights reserved. David Publishing Company holds the exclusive copyright of all the contents of this journal. In accordance with the international convention, no part of this journal may be reproduced or transmitted by any media or publishing organs (including various websites) without the written permission of the copyright holder. Otherwise, any conduct would be considered as the violation of the copyright. The contents of this journal are available for any citation. However, all the citations should be clearly indicated with the title of this journal, serial number and the name of the author. Subscription Information: Price: Print $360 (per year) Online $300 (per year) Print and Online $560 (per year) David Publishing Company 1840 Industrial Drive, Suite 160, Libertyville, Illinois 60048 Tel: 1-847-281-9826. Fax: 1-847-281-9855 E-mail: [email protected]

Journal of Energy and Power Engineering Volume 4, Number 8, August 2010 (Serial Number 33)

Contents Clean and Sustainable Energy 1

Performances of Flat Plate Pulsating Heat Pipes W. Qu and B. Yang

9

Prediction of Boiler Drum Pressure and Steam Flow Rate Using Artificial Neural Network A.T. Pise, S.D. Londhe and U.V. Awasarmol

16

Design and Development of a Laboratory Scale Biomass Gasifier S.J. Ojolo and J.I. Orisaleye

24

Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator K. Oshima and Y. Uchiyama

Power and Electronic System 32

A Super Low Power CMOS Receiver for High Resolution Epi-retinal Prosthesis J.W. Yang, N. Tran, S. Bai, D.C. Ng, M. Halpern and E. Skafidas

40

ANN Based Performance Analysis of UPFC in Power Flow Control S.K. Srivastava and S.N. Singh

48

Spectral Calculation of the Output Voltage of an Inverter with Bipolar Pulse Width Modulation I. Mrčela, V. Šunde and Z. Benčić

57

The Energy Balance of the Electro-Hydraulic Linear Actuation System M. Herranen, K. Huhtala and M. Vilenius

August 2010, Volume 4, No.8 (Serial No.33) Journal of Energy and Power Engineering, ISSN 1934-8975, USA

Performances of Flat Plate Pulsating Heat Pipes W. Qu and B. Yang Institute of Engineering Thermophysics, Chinese Academy of Sciences, Beijing 100190, China Received: February 03, 2010 / Accepted: March 24, 2010 / Published: August 31, 2010. Abstract: Flat plate pulsating heat pipe is useful for hot spot heat spreader. Two kinds of flat plate spreader of pulsating heat pipe are designed, fabricated and experimented. For the embedded circular capillary type, the transferred heat flux could reach 32 W/cm2, the smallest thermal resistance for acetone, methanol and FC-72 were respectively 0.50, 0.57 and 0.40 ℃/W. While for the square capillary type, the transferred heat flux could reach 26 W/cm2, the equivalent thermal conductivity could reach 3211 W/(m·℃). There are ranges of optimal transferred power and filling ratio for different working liquid. If the transferred power is constant, changing the heating area and the place has little effects on the performance. Key words: Pulsating heat pipe, heat spreader, the embedded circular capillary type, the square capillary type, heat flux.

1. Introduction The power of electronics is continuously increased year by year. Thermal management is one of the most critical technologies in electronic products, and sometimes the cooling technology directly influences the cost and reliability of the products [1-3]. On the one hand, the previous heat transport technology should be improved, on the other hand, the new cooling technology also should be focused [4]. The temperature of operating electronics should be controlled to a certain level, then the electronics can work normally and have the reliability and longer life. Generally, the normal temperature range of the operating electronics should be between -5～65 ℃. If the temperature of local or whole reaches a higher level, such as 90℃, then the operating stability and reliability will decrease a lot. Such as, for a single semiconductor element, if the temperature is increased 10 ℃ higher than that of the normal operating, then the reliability of the system will decrease 50% [5]. Additionally, the nonuniformity of temperature distribution can lead to the thermal stress and further Corresponding author: W. Qu, professor, research fields: heat pipe, phase change heat transfer, enhanced heat transfer. E-mail: [email protected].

the inner distortion of electronics, this can make the contact thermal resistance increased, and the heat transfer is becoming worse. The electronics could become failed by fatigue, split or permanent distortion [6, 7]. The stability of the device and the system is affected by the heat transfer element. Some data show that some failures of the electronics are caused by the over temperature of the local or whole electronics, the ratio can be 20% or so. The electronics is generally limited to a restricted space, the common cooling technology can not satisfy some electronics cooling demands. There is a tendency to pack multiple bare chips into one module. One possibility to extend the limits is to mount the chips directly on flat mini-or micro-heat pipe arrays [8]. Such heat pipes typically consist of a set of fine grooves which are machined into flat structures. However, the fabricating of true micro heat pipe arrays requires higher precision tools and needs very complex processes. And they have smaller total areas and are thicker, so they are not convenient for industry. The inner capillary structure of the heat pipe could have a new design, also the shape could be designed specially [9, 10]. Pulsating heat pipe (PHP) is a new type heat pipe. Flat plate heat spreader of pulsating heat

2

Performances of Flat Plate Pulsating Heat Pipes

pipe has a promising use in the field of electronics cooling, especially for the hot spot situation [11]. The objective of this study is the realization of two dimensional heat spreader of pulsating heat pipe. Firstly, two kinds of heat spreader, the embedded circular capillary and the square capillary types, are designed and fabricated. Secondly, the performances of them are experimented. Finally, the results were presented and analyzed, and concluded. Larger and thinner heat spreader of pulsating heat pipe is a new direction for hot spot cooling of electronics. Two different kinds of pulsating heat pipe spreader are designed and fabricated. One is the embedded circular capillary type, the other is the square capillary type.

2. Heat Spreader Structures of Pulsating Heat Pipe 2.1 The PHP Spreader of Embedded Circular Capillary Type The PHP spreader of embedded circular type consists of two parts. One is the thin copper tube, the size is Ф2×0.5 mm, which is bended many times to form a looped PHP, it has 12 turns and 24 straight sections. The other part is two thin aluminum plates, each of them is fabricated as 24 axial semi-circular grooves. The bended tube is embedded between the grooves of the two plates. Between the tube and the groove, the two plates, the conduction grease is daubed to decrease the thermal resistance. The two parts are coupled by several mini bolts as one PHP spreader, the size of the coupled aluminum plate is 200 mm×60 mm×4 mm. The distance of the two middle tubes is 4 mm, while for the other two neighbor tubes, they are contacted. Fig. 1 shows the picture of the PHP spreader of embedded circular capillary type. 2.2 The PHP Spreader of the Square Capillary Type The second PHP spreader is designed as 22 square capillary passages, aluminum looped pulsating heat pipe. The cross-section of the square capillary is 1.5

Fig. 1 The PHP spreader of embedded circular capillary type.

Fig. 2 The PHP spreader of the square capillary type.

mm×1.5 mm. The size of the spreader is 250 mm×60 mm×4 mm. Fig. 2 illustrates the picture of the spreader.

3. Experimental Setup The experimental setup of the heat spreader is schematically shown in Fig. 3. The heat spreader is heated and cooled on top side, and the heater can be moved to the lower side. The heat is applied by one copper heater of high heat flux with four layers of insulations. The condenser is cooled by the pumped water circulation to the jacket. The contact thermal resistance is decreased by one thin layer heat conduction silicon grease both between the heater, the cooling jacket and the heat spreader. The real heating power is the total minus that of the dissipation to ambience. The other parts of the heat spreader are wrapped by the insulation cotton, so the heating power to the thin flat plate heat pipe (TFPHP) could be calculated accurately. The heat spreader could be heated by 20×20 mm2 or 20×10 mm2 heater. The area of the cooling jacket is 100×60 mm2. The TFPHP includes the evaporator of 20 mm, the adiabatic section of 120 mm and the condenser of 120 mm, where two, three and two thermocouples of 0.2 mm diameter nickel-chromium and nickel-aluminum are set along the length. The temperature data could be got by one HP digital multimeter, also they could be displayed by the computer.

3

Performances of Flat Plate Pulsating Heat Pipes Heating area

Cooling area

Fig. 4 Heating and cooling areas, position thermocouples of the embedded heat spreader. 1-computer; 2-multi channel data acquisition system; 3-heat spreader; 4-high heat flux heater; 5-cooling jacket; 6-flow meter; 7-thermal insulation; 8-water pump; 9,10-cooling water tank

Heating area

of

Cooling area

Fig. 3 Experimental setup of the two heat spreaders.

As shown in Fig. 4, for the embedded heat spreader, five thermocouples of K type (NiCr-NiAl) are set along the length, opposite to the heating and cooling side, while at the front and bottom of the spreader, one thermocouple each are placed. For the square capillary heat spreader, six thermocouples of K type (NiCr-NiAl) are set along the length, opposite to the heating and cooling side. The heating and cooling areas, the positions of thermocouple are also shown in Figs. 4 and 5. The heads of the thermocouples are adhered to the plate or tube walls by adhesive, and for each spreader, one plate is used for fixing the thermocouples. The thickness of ach spreader is only 4 mm, the thermocouples can illustrate the temperature distribution of the spreader.

4. Results and Discussion The performances of the spreaders can be determined by the equivalent thermal conductivity and the thermal resistance as: keff =

Q × leff

A × (Te − Tc )

R = (Te − Tc ) Q

(1) (2)

The operating temperature and the temperature difference of the spreaders can be expressed as: Twork = (Te Le + Ta La + Tc Lc ) L

(3)

(4) ΔT = Te − Tc Where, Q is the net heating power, equals to the total power minus the heat leakage, which is estimated by

Fig. 5 Heating and cooling areas, position of thermocouples of the square capillary heat spreader.

temperature difference between the heater wall and the ambience. leff is the distance between the center positions of the heating section and cooling section. Te and Tc are mean temperature of the heating section and the cooling section respectively. A is the cross-section area of the spreader. For the embedded spreader, there are three ways for the heat transfer as shown in Fig. 6, one is by the base aluminum plate, the second is by the capillary tube wall, and the third is by the pulsating heat pipe operation. The performance of the spreader can be demonstrated by Rphp,effect, while Rempty,eff stands for the effects of the spreader case, they can be expressed as: A A (5) kempty ,eff = k Al × Al + kCu × Cu Aw Aw 1 1 1 (6) = + Roverall R php ,effect Rempty ,eff By calculation, the equivalent thermal conductivity of the empty case is 245.42 K/W. For the square capillary heat spreader, there are two ways for the heat transfer, one is by the base aluminum case, the other is by the pulsating heat pipe operation. 1 1 1 = + Roverall Rphp ,effect Rempty

(7)

Where, Rphp,effect reflects the PHP performance, Rempty is the equivalent thermal conductivity of the empty case.

4

Performances of Flat Plate Pulsating Heat Pipes 2

Table 1 Filling rate of the two heat spreaders. Type The embedded The square capillary

Working fluid Acetone Methanol FC-72

Filling rate (%) 36, 50 38 38, 52, 60, 67, 80.6, 88

Acetone

20.9, 50

For the embedded heat spreader, the working fluids are selected as acetone, methanol and FC-72, while for the square capillary type, only acetone is selected. The corresponding filling rates are shown in Table 1. The heating area can be chosen as 20×20 mm2 and 10×20 mm2. The tilt angle is as 90º, 60º, 30º and 0o. For the embedded and square capillary types, the middle temperature of the heater is controlled below 120 ℃ and 100 ℃ respectively. 4.1 The Embedded Type Heat Spreader For the embedded type heat spreader, the angle between the length direction and the horizon is expressed as α, while the width direction and the horizon is set as β. 4.1.1 The Effects of Heat Flux Fig. 7 shows the effects of transferred power and heat flux to the thermal conductivity and thermal resistance of the embedded type sample. The working fluid is acetone. The filling rate is 36%. The heating area is 4 cm2, upper surface heated. The tilt angle of the spreader is α=90º, bottom heated. The thermal conductivity will increase with the transferred power and heat flux, then decrease a little. The maximum keff of the spreader can reach 1,200 W/(m·K) or so, the corresponding thermal resistance is 0.52 K/W, and the heat flux is 10.5 W/cm2, the working temperature is only 42 ℃. The better transferred power range demonstrates as from 28 W to 68 W, the heat flux should be between 7-17 W/cm2.

4

6

Heat flux q (W/cm ) 8 10 12 14

16

18

1200

1.4

1050

1.2 Acetone Fr=36% keff

900 750

R

1.0 0.8

600 0.6 450

8

16

24 32 40 48 56 Transferred power Q (W)

64

Thermal resistance R (K/W)

Fig. 6 The parallel connection of the thermal resistances of the embedded heat spreader.

Thermal conductivity keff (W/(m.K))

2

72

Fig. 7 Thermal conductivity and thermal resistance as a function of transferred power and heat flux.

4.1.2 The Influences of the Spreader Tilt Angle For the spreader, there are two effects of the tilt angle to the operation. The tilt angle or the gravity can change the distribution of liquid plugs in the circular capillary tubes, when the tilt angle is β=0º, α>0º. The bottom capillary tubes can be wetted much more than those of the top for the same transferred power, especially for the direct position, α=90º. On the other hand, the gravity can influence the flow forces to the liquid plugs. During operation, the liquid distribution along each capillary tube is different, the gravity can provide more driving force to the operation. Fig. 8 illustrates the experimental result summation of the influences of the tilt angle. The width angle is β=0º. For different tilt angle, the minimum thermal resistance corresponds to different transferred power or heat flux. As shown in Table 2, the spreader can transfer larger power horizontally, however, the thermal resistance is a little bigger. The tilt angle can realize smaller thermal resistance, however, the power transferred should be a little smaller. 4.1.3 The Effects of Filling Rate The filling rate of the spreader is defined as the ratio of the filling liquid volume to that of the total capillary space. For pulsating heat pipe, the recommended filling rate of startup and successful operation is 20%~80%. Generally, pulsating heat pipe has an optimal filling rate, this can make thermal resistance the smallest, it

5


4

6

Heat flux q (W/cm

8

10

12

)

2

14

16

18

1.8

Acetone, Fr=36%, β=0°

1.4

α=0°



1.6

2

α=30°

1.2

α=60° α=90°

1.0 0.8 0.6 0.4

6

8

16


64

72

26

Fr=36% Fr=50%

1.4 1.2 1.0 0.8

24

32

Transferred power (W) 65 44 44 44

Thermal resistance (K/W) 0.68 0.56 0.50 0.50

depends on the transferred power, number of turns, species of working fluid, etc.. For some conditions, the filling rate can change the thermal resistance a lot, as shown in Fig. 9, when the working fluid is acetone. The performance of 36% filling rate spreader is much better than that of 50%. If the working fluid is changed as FC-72, for the 38% filling rate of the spreader, the thermal resistance is the biggest, compared with those of 52%, 60%, 67% and 80.6% as shown in Fig.10. Fig. 10 shows the results of several filling rates of FC-72. Compared the two results of Figs. 9 and 10, the best performance of the spreader corresponds to 36% filling rate for acetone and 67% filling rate for FC-72. From Figs. 9 and 10, the best filling rates are also shown. For the same center temperature of the heating section, if the filling rate is selected differently, the biggest transferred power changes. When the temperature of the heating section is 100 ℃, the transferred power can be over 100 W at the best filling rate 67%, the corresponding thermal resistance is 0.4 K/W. If the spreader is put horizontally, the biggest

40

48 56 64 72 80 88 Transferred power Q (W)

96 104

Fig. 9 Thermal resistance of two filling rates when the working fluid is acetone. 2

8

Heat flux q (W/cm ) 12 16 20 24

28

32

2.0 Thermal resistance R (K/W)

0 30 60 90

24

Acetone, α=90°, β=0°

1.6

Table 2 Filling rate of the two heat spreader.

(º)

22

0.6 8

Fig. 8 Thermal resistance versus transferred power at different tilt angle.

Tilt angle

10

Heat flux q (W/cm ) 12 14 16 18 20

FC-72, α=90°, β=0°

1.8

Fr=38% Fr=52% Fr=60% Fr=67% Fr=80.6%

1.6 1.4 1.2 1.0 0.8 0.6 0.4 32

48 64 80 96 Transferred power Q (W)

112

128

Fig. 10 Thermal resistance of five filling rates when the working fluid is FC-72.

transferred power is only 57 W at the best filling rate is 52%, the corresponding thermal resistance is 0.9 K/W. 4.1.4 The Influences of Working Fluid Species For different working fluid, the performances of the spreader sample can change a lot. In this paper, the working fluids of methanol, acetone and FC-72 are considered and compared. For nearly the same filling rate of 38%, the thermal resistance of the spreader sample is shown and compared as in Fig. 11. For one same power, acetone is the best for the working fluid selection. FC-72 and methanol are almost the same for the performance, however, the methanol can make the spreader sample transfer much more power and heat flux, the biggest heat flux can be 26 W/cm2 or so, the corresponding

6


6

resistance R (K/W)

Heat flux q (W/cm ) 10 12 14 16 18 20 22 24 26 α=90°, β =0°

1.6

Thermal

8

Methanol, Fr=38% Acetone, Fr=36% FC-72, Fr=38%

1.4 1.2 1.0 0.8 0.6 24

0.40

3300

keff R

3000 2700

0.30

2400

0.25

2100 0.20 1800 30

32

40

48 56 64 72 80 88 Transferred power Q (W)

96 104

0.35

40

50

60

70

80

90


4


2

2

0.15 100 110

Transferred power Q (W)

Fig. 11 Thermal resistances of different working fluid.

Fig. 12 Thermal conductivity and thermal resistance as a function of the transferred power.

thermal resistance is also the smallest, as 0.57 K/W. The results tell that, for 1 mm capillary, at 38% filling rate, if the transferred power is smaller, acetone is a good selection; if the transferred power is bigger, methanol is a good selection.

4.2.2 The Influences of the Spreader Tilt Angle For the square capillary spreader, at different tilt angle as α=90º 、60º 、30º , the results of thermal conductivity versus the transferred power are shown in Fig. 13.

4.2 The Square Capillary Heat Spreader For the angular capillary pulsating heat pipe, due to the angular action to the liquid, there need not much liquid to operate the working fluid. The filling rate is experimented to be smaller. 4.2.1 The Effects of Transferred Power For the spreader sample of square capillary type, aluminum is taken as the case, the filling rate is 20.9%, the working fluid is acetone. The tilt angle α=90º, and the mode is the bottom heating. The results of the transferred power to the thermal conductivity and the thermal resistance at tilt angle α=90º are shown in Fig. 12. With the transferred power, the thermal conductivity will increase at first, to the maximum, then decrease. The thermal resistance is the opposite, decrease to the minimum, then increase. The biggest experimental transferred power reaches 102 W, the heat flux is 25.7 W/cm2. When the thermal conductivity increases to the maximum, the equivalent thermal conductivity keff is over 3,200 W/(m·K), it is 16 times that of the spreader case. The corresponding thermal resistance is 0.195 K/W, the heat flux is 15.8 W/cm2. The working temperature is only 39 ℃.

With the transferred power, the performances all show one maximum thermal conductivity, however, the best performance occurs at different transferred power. The tilt angle 90º corresponds to the transferred power, thermal conductivity and thermal resistance as 63 W, 3,211 W/(m·K) and 0.195 K/W; For 60º tilt angle, they are 72.64 W, 2,476 W/(m·K) and 0.25 K/W; For 30º tilt angle, they corresponds to 61 W, 1,766 W/(m·K) and 0.35 K/W. This spreader also shows good performance if it is bottom heating, that is, the gravity assists to operate. 4.2.3 The Effects of Filling Rate For two filling rate as 50% and 20.9% of the spreader sample, the results of the thermal resistance are demonstrated in Fig. 14. For the same transferred power, the 20.9% filling rate spreader has the smallest thermal resistance as 0.19 K/W. The minimum thermal resistance of the 50% filling rate is higher as 0.77 K/W. The tentative explanation is, the higher filling rate can make the liquid bridge easier. The capillary hysteresis can increase the flow resistance of the liquid plugs. For the angular capillary pulsating heat pipe, there need not much more liquid compared with those of the circular. The angular

7


2

Heat flux q (W/cm )

10

0

α=90 0 α=60 0 α=30

2800 2400 2000 1600 1200

30

40

50

60

70

80

90

24

26

0

0

Methanol, Fr=38%, α=90 , β=0 2

Ae=2cm

1.2

2

Ae=4cm

1.0 0.8 0.6

100

40

Transferred power Q (W)

Fig. 13 Thermal conductivity versus power at different tilt angle.

12

1.4 Thermal resistance R (K/W)


7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 3200

Heat flux q (W/cm ) 14 16 18 20 22

48


96

104

Fig. 15 The effects of heating area to the thermal resistance. Heat flux q (W/cm ) 2

2

20

25

0.95

6

8

10

12

14

16 0

0

Acetone, Fr=36%, α=0 , β=0 0

1.6

Fr=50%, α=90 0 Fr=50%, α=60 0 Fr=20.9%, α=90 0 Fr=20.9%, α=60

1.4 1.2 1.0 0.8 0.6 0.4



1.8

5

Heat flux q (W/cm ) 10 15

0.90 0.85 0.80 0.75 0.70 0.65 24

0.2 20

40 60 80 100 Transferred power Q (W) Fig. 14 Thermal resistance versus transferred power and heat flux at different filling rate and tilt angle.

field can make the spreader work as two mechanisms, one is micro heat pipe, the other is pulsating heat pipe. 4.3 Influences of Heating Area and Position 4.3.1 The Effects of Heating Area For the embedded type spreader sample, the results of the heating area influences are shown in Fig. 15, when the working fluid is methanol, the filling rate is 38%, and the mode is bottom heating. The two heating areas are selected as 2 cm2 and 4 cm2. From the results, for most transferred power, the smaller heating area corresponds to the smaller thermal resistance. If the heating area is 2 cm2, and the

upper side lower side

32

40 48 56 Transferred power Q (W)

64

Fig. 16 The effects of upper or lower side heating to the thermal resistance.

transferred power is 98 W, the thermal resistance will be 0.64 K/W. While if the heating area is 4 cm2, when transferred power is 103 W, the thermal resistance will be 0.58 K/W. 4.3.2 The Effects of Upper or Lower Side Heating For the embedded spreader sample, when the mode is bottom heating, the heating area is 4 cm2, the filling rate is 36%, the working fluid is acetone, the results of the heating position to the performance are shown in Fig. 16. As shown in Fig. 16, if the transferred power is smaller, such as less than 50 W, the thermal resistance of the lower side heating has smaller thermal resistance. 39W transferred power corresponds to the smallest

8


thermal resistance as 0.69 K/W. While at the higher transferred power, the upper side heating has better performance. The transferred power of 65 W corresponds to the smallest thermal resistance of 0.68 K/W. As a whole, the influences of upper or lower heating position to the thermal resistance are not significant.

5. Conclusions Two different types of pulsating heat pipe heat spreader are designed and fabricated. They are the embedded type and the square capillary. For the embedded sample, its highest heat flux can reach 32 W/cm2. There are optimal heat loads and optimal filling ratios for different kinds of the working liquid. For acetone and FC-72, the optimal filling rates are 36% and 67% respectively. The tilt angle of PHP influences the performance a lot, the bottom heating is better. The smallest thermal resistances are 0.5, 0.57 and 0.4 K/W respectively for acetone, methanol and FC-72 at 90º tilt angle. The heating area and its position have little effects on the performance of PHP. For square capillary sample, its highest heat flux could reach 26 W/cm2. When the filling ratio is 20.9%, the heat spreader showed better performance, the thermal resistance is 0.195 K/W, the smallest. There are optimal heat loads and optimal filling rates for different tile angles. For the tilt angles of 90º, 60º and 30º, their thermal resistance could be 0.195, 0.25 and 0.35 K/W respectively.

Acknowledgments This research is financially supported by NSFC with Grant No. 50676096.

References [1]

R.C. Chu, R.E. Simons, G.M. Chrysler, Experimental investigation of an enhanced thermosyphon heat loop for cooling of a high performance electronics module, in: Proc. of l5th IEEE Semi-Thermo Symposium, San Diego, Canada, 1999. [2] P. Meena, S. Rittidech, N. Poomsa-ad, Application of closed-loop oscillating heat-pipe with check valves (CLOHP/CV) air-preheater for reduced relative-humidity in drying systems, Applied Energy 84 (2007) 553-556. [3] Z. Zuo, M. North, Miniature high heat flux heat pipes for cooling of electronics, in: Proc. of SEE2000, pp. 573-579. [4] T. Wong, P. Terdtoon, P. Sakulchangsatjatai, Operation moldeling of closed-end and closed-loop oscillating heat pipes at normal operating condition, Applied Thermal Engineering 24 (7) (2004) 995-1008. [5] W. Qu, T.Z. Ma, Experimental Investigation on Flow and Heat Transfer of a Pulsating Heat Pipe. Proceedings of 12th Int. Heat Pipe Conference, Moscow, Russia, 2002, 226-231. [6] S. Khandekar, P. Charoensawan, M. Groll, P. Terdtoon, Closed loop pulsating heat pipes Part B: visualization and semi-empirical modeling, Applied Thermal Engineering 23 (2003) 2021-2033. [7] P. Charoensawan, Nondimensional correlation to predict the thermal performance of horizontal closed-loop pulsating heat pipe, in: 14th Int. Heat Pipe Conference, Florianopolis, Brazil, Apr. 22-27, 2007. [8] P. Charoensawan, S. Khandekar, M. Groll, Closed looped pulsating heat pipes: Part A: parametric experimental investigation, Applied Thermal Engineering 23 (2003) 2009-2020. [9] A. Bensalem, C. Romestant, A. Alexandre, Y. Bertin, Experimental investigation of pulsating heat pipes, in: 14th Int. Heat Pipe Conference, Florianopolis, Brazil, Apr. 22-27, 2007. [10] S. Khandekar, N. Dollinger, M. Groll, Understanding operational regimes of closed loop pulsating heat pipes: an experimental study, Applied Thermal Engineering 23 (2003) 707-719. [11] Khandekar, M. Groll, An insight into thermohydrodynamic coupling in closed loop pulsating heat pipes, Int. Journal of Thermal Sciences 43 (2004) 13-20.


Prediction of Boiler Drum Pressure and Steam Flow Rate Using Artificial Neural Network A.T. Pise1, S.D. Londhe2 and U.V. Awasarmol3 1. Department of Mechanical Engineering, Government College of Engineering, Karad, Maharashtra 415124, India 2. Department of Mechanical Engineering, Government College of Engineering, Chandrapur, Maharashtra 442401, India 3. Department of Mechanical Engineering, Army Institute of Technology, Pune, Maharashtra 411015, India Received: March 01, 2010 / Accepted: April 14, 2010 / Published: August 31, 2010. Abstract: Numerical simulation of complex systems and components by computers is a fundamental phase of any modern engineering activity. The traditional methods of simulation typically entail long, iterative processes which lead to large simulation times, often exceeding transient real time. Artificial neural networks (ANNs) may be advantageous in this context, the main advantage being the speed of computation, the capability of generalizing from the few examples, robustness to noisy and partially incomplete data and the capability of performing empirical input-output mapping without complete knowledge of underlying physics. In this paper, the simulation of steam generator is considered as an example to show the potentialities of this tool. The data required for training and testing the ANN is taken from the steam generator at Abott Power Plant, Champaign (USA). The total number of samples is 9600 which are taken at a sampling time of three seconds. The performance of boiler (drum pressure, steam flow rate) has been verified and tested using ANN, under the changes in fuel flow rate, air flow rate and load disturbance. Using ANN, input-output mapping is done and it is observed that ANN allows a good reproduction of non-linear behaviors of inputs and outputs. Key words: Boiler, artificial neural network, steam flow rate, drum pressure.

1. Introduction Boiler and turbine system is the complex industrial process with high nonlinearity, uncertainty and load disturbance. A lot of efforts have been made on modeling and control of steam generators from 1970s. The plant can be represented by a series of nonlinear equations describing energy and mass balance. In recent years, artificial intelligence has been incorporated in the modeling of many applications like steam generator. The reasons for extensive attention to the boiler control is to improve plant efficiency, life time, load following capability and environmental protection etc. [1]. In the boiler system, the drum steam pressure and steam flow rates are the important factors affecting the Corresponding author: A.T. Pise, professor, research fields: heat and mass transfer, refrigeration and air conditioning. E-mail: [email protected].

combustion process. A boiler system must maintain the desired steam pressure and the steam flow rate at the outlet of drum because these are the key parameters to control the output of the turbine. The advances in sensor and control systems are required in order to meet the increasingly stringent standards on emission. The extensive work has been done by Allen et al. [2], to design and propose imaging combustion system. The working steam pressure also affects the quality of steam which is described in terms of amount of Na, SiO2 and CO2 dissolved into steam. The dissolved substances in the boiler water pass into steam which therefore gets contaminated [3]. Several research papers have been presented for linear, non-linear mathematical models of different steam generator. Adam and Marchetti [4] developed two separate non-linear mathematical models for simulation of boiler, one for evaporation in vertical

10

Prediction of Boiler Drum Pressure and Steam Flow Rate Using Artificial Neural Network

tubes and other for phase separation in the steam drum. Tarasiewicz, and Radziszewski [5] modeled an oil fired boiler mathematically in two distinct subsystems, one for thermo-chemical process of combustion and other for thermodynamic process of steam generation. Heat transfer phenomenon in both combustion chamber and in main boiler tubes was coupled. In the mathematical model described by Lawa et al. [6], the simulation of boiler was carried out on the basis of static and dynamic data obtained from a small scale plant. They reported the validation of simulated data with the actual data. Here the main problem came from the improper balance between steam flow rate and feed water flow rate; hence it is very important that the corresponding measurement errors corrected. It was also assumed that the feed water flow was errorless due to which the results were not accurate. Alatiqi and Meziou [7] formulated a dynamic mathematical model in which the transient response of water level in steam drum due to changes in feed water flow rates were simulated. In these above mathematical models, numbers of complicated mathematical formulae were suggested for different parts of the boiler. Their solutions are difficult as well as time consuming. Also in the mathematical models, various assumptions are made, which develop inaccuracy in the results and thus the mathematical models does not meet the real processes at different steady states accurately. In the artificial neural network (ANN) model presented here, the care of all above parameters is taken and ANN is proved to be the best tool for the simulation of boiler operations. Also this ANN model, unlike the mathematical model can be applied to the entire boiler, i.e. separate ANN model is not required for the different parts of the boiler. Knowing the potentials of ANN, the output parameters discussed here are strongly dependant on the input parameters such as air flow rate, load disturbances, fuel flow rates which were not considered in the simulation by the Liu et al. [1], and Masini et al. [8]. This paper aims to predict the steam drum pressure and steam flow rate in the boiler with help of simulation software

‘Neuro-Solutions’. 2. Artificial Neural Network System (ANN) In this work, ANN is used to perform the task of prediction of drum pressure and steam flow rate in boiler. Back Propagation Algorithm with three layered network is proved to be capable of mapping any kind of non linear data with the required accuracy. Design of neural network involves following aspects: number of input layer neurons, output layer neurons, hidden layers and hidden neurons and learning strategy. 2.1 Number of Layers and Neurons Input and output layers and neurons: Since there are three independent variables (i.e., air flow, fuel and disturbance load) as inputs to the network, the number of input neurons will be three. In a similar fashion, the number of output neurons will be equal to the response variables of the network which are two here, i.e., steam pressure and steam flow rate. Hidden layers and hidden neurons: The number of hidden layers plays a very important role in the accuracy and efficiency of the network. It is proved that a three layered network, that is one input layer, one hidden layer and one output layer, can approximate any kind of non-linearity with the required accuracy, so here also the network is three layered [9]. Here the evaluation version of ‘Neuro-Solutions’ software is used for which the maximum number of neurons that can be used is fifty. The neural network is tested for the different values of neurons in hidden layer. It has been observed that the network performs well when fifty neurons are used in hidden layer. 2.2 Learning Strategy The neural network needs to be trained before it is employed in actual practice. The training strategy is an important factor in designing the network. Usually, back propagation learning algorithm provides computationally efficient method for training the multi layered perceptron (MLP) networks. In this algorithm,


11

the error is back propagated to change the parameters of network. This is called learning of the network. The learning can be done in either pattern mode or in batch mode. In pattern mode, the parameters updating is performed after the presentation of each training sample whereas in batch mode, the parameters updating is performed after the presentation of all the training examples that constitute an epoch. Literature reveals that multi-layer perceptron network with back propagation algorithm (generalized delta rule) is used input-output mapping, i.e., in regression model. Back propagation algorithm: In the present system, back propagation algorithm used is depicted in the flow chart (Fig. 1). Linear, sigmoid as well as hyperbolic tangent functions are used to estimate the response of neurons in hidden layer and output layer. The error in the output signal is back propagated to calculate the new weights.

3. Software Details The data can be processed by preparing ANN software program in C language. Some standard codes are also available in MatLab to construct ANN model. Here the data used is huge, so due to time constraints, the data is processed and simulated in ‘NeuroSolutions’ software. The program is divided into mainly three modules: input module, computation module and output module (Fig. 2). In input module, the program gets inputs from file which contains the training patterns. Neural network is initialized in this module. The data read from this input file is used in the computation module. ‘Neuro-Solutions’ software normalizes as well as randomizes the patterns before processing in the computation module. Computation module is the core module of the program which consists of functions to carry out the tasks such as error calculations, back propagation algorithm and training. The parameters are adjusted according to error calculated and finally neural network get trained up to predefined accuracy. The output module does the front end tasks such as calculation of output for the testing data

Fig. 1 Flow chart for ANN model.

Fig. 2 ANN program details.

and print the output on screen or store in a file. It consists of the inputs and the output corresponding to those inputs. It also has the facility to either store or print the whole input training data as a file for future reference.

4. Training of ANN 4.1 Normalization of Input Data It is recommended that the input data used to train the

12


neural network should be in normalized form that is each input is to be converted to a value between 0 and 1. The software normalizes the data on its own. However, there are a number of ways to normalize the input data, both using linear and non linear methods [10]. 4.2 Criteria for Stopping ANN Training The back propagation algorithm in general can not be shown to converge. Also, there are not well defined criteria to stop the training. To formulate such a criterion, one has to think of the unique properties of the minimum. Let W* denote a minimum, local or global. A necessary condition for W* to be minimum is that the gradient vector of the error surface with respect to the weight must be zero at W equal to W*.They are certain criteria which are given below [11, 12]. Since the problem in this system is of ‘function approximation’ type, the following stopping criterion is used. “The back propagation is considered to have converged if the average squared error is less than the square of the given permissible error.” 4.3 Momentum Term (α) Back propagation leads the weights in a neural network to a local minimum of the Mean Square error (MSE), possibly substantially different from the global minimum that corresponds to the best choice of weights. This problem can be bothersome if the error surface (plotting MSE against network weights) is highly uneven or jagged as shown in Fig. 3 [12]. This problem can be avoided by making the weight changes depend on the average gradient of MSE in a small region rather than the precise gradient at a point. Averaging in a small neighborhood can allow the network weights to be modified in the general direction of MSE decrease, without getting stuck in some local minima. 4.4 Generation of Training and Testing Data The training of an ANN to simulate the dynamic response of steam generator requires the availability of

complete data set concerning the different transient examples. In principle, it would seem better to take the historical data from a real steam generator, than the simulated code, which contains some physical and mathematical simplifications as well as uncertainties due to many assumptions made [2]. In this paper the actual data is taken from the Abbott Power Plant, Champaign. The data contains 9,600 samples taken at a sampling time of 3 seconds of all the data samples, 60% data is used for training the network, 15% data is used for cross validation and remaining 25% data is kept aside for testing purpose.

5. Selection of Parameters From the literature survey, importance of all the parameters is studied. The important parameters which affect the performance of the boiler are air flow rate, fuel flow rate, drum water level and load disturbance. The steam generated in the boiler is given to turbine, hence the steam pressure and steam flow rate are of prime importance, which are considered here as response parameters. Input parameters: Air flow rate, fuel flow rate and

Fig. 3 Momentum term. Table 1 Optimum design parameters with the minimum error. S.N. Parameter Value Error 1 No. of input layer neurons 3 2 No. of output layer neurons 2 3 No. of hidden layer neurons 50 4 Transfer function of hidden layer Linear sigmoid 0.0079 5 Transfer function of output layer Sigmoid 6 Momentum of hidden layer 0.7 7 Momentum of output layer 0.85 8 Epoch 1000 9 Weight update Batch wise

13


load disturbance. Output parameters: Drum pressure, steam flow rate.

6. Results and Discussions The ANN model which is arrived at is the outcome of much efforts and continuous process. The parameters which were considered to design this model are: (1) number of hidden layers (0 to 3) (2) number of processing elements in the hidden layers (5 to 50 in the step of 5) (3) transfer function of hidden layer (5 numbers, linear, sigmoid, linear sigmoid, tanh and linear tanh) (4) transfer function of the output layer (5 numbers, linear, sigmoid, linear sigmoid, tanh and linear tanh) (5) momentum term (0.5 to 0.9 in the step of 0.1) (6) number of epochs (100 to 1,500 in the step of 50) This optimum model has been reached by considering almost all possible permutations and combinations of different values of the above parameters. The ANN model optimum design parameters with minimum error are shown in Table 1. 6.1 Training of the Network 60% of the total data which is meant for training is used to train the neural network. The neural network is trained by using all possible combinations of the various parameters such as activation functions, epochs, momentum etc.. Once the data is given to the network, it starts learning from the data. From the training data the network generalizes the problem and does not remember it. The training is to be terminated when the cross validation error starts increasing. An ideal training is one that converges fast to reduce the error to zero. There is a significant effect of momentum on the learning of the network which in turn affects the output of the network. This effect is shown in Figs. 4 a, b, and c and is summarized in the Table 2. 6.2 Testing of the Data 25% of the total data is kept aside for testing purpose. The data is tested in neuro-solutions testing Wizard. On

Table 2 Effect of momentum on error. Hidden layer Output layer Transfer function Linear sigmoid Sigmoid 0.8 0.85 Momentum 0.7 0.9

Error 0.08410 0.007964 0.008349

a trained network as designed shown in Table 2 for which the error is 0.007964, the test data is tested and the actual and simulated output parameters are compared on the graphs. It is evident from the Fig. 5 that with linear-linear activation functions, the learning process is poor Thus Fig. 5 shows a large deviation in the actual and simulated results. With linear sigmoid-sigmoid functions and with an output layer momentum of 0.85 and hidden layer momentum of 0.7, the training is better (Fig. 4) and the results are also promising which are shown in the Fig. 6. The output parameters discussed here are strongly dependant on the input parameters such as air flow rate, load disturbances, fuel flow rates which were not considered in the simulation by the above researchers. These parameters are considered in this ANN model presented here. The data were simulated and tested with the actual data; results show the better agreement between simulated and actual data. An extensive work has been done to find the mathematical models for the different parts of the boiler. However, a very few works using ANN have been reported. Moreover, the number of input parameters is not adequate as discussed in the literature survey. The ANN is found to be the best tool for the solution of linear as well as non-linear problems. However, its contribution and use reported in the literature is not so significant. This work takes into account number of output dependant parameters and may prove to be the useful addition to prove the applicability of ANN for complex problems. There are many industries where boilers are used for various applications of process heating and power production. Such a model developed can be used in the real life steam generator for online monitoring and

14


(a) α = 0.8

(b) α = 0.85

(c) α = 0.8 Fig. 4 Effect of variation of momentum term on training process. 600

600

500

500 400 300 200

Des Drum_p

400

Des Steam_

300

Out Drum_p

200

Out Steam_

100

100

Des Drum_p Des Steam_ Out Drum_p Out Steam_

0

0

1 83 165 247 329 411 493 575 657

1 83 165 247 329 411 493 575 657

Fig. 5 Actual and simulated results with linear-linear functions.

controlling of the output parameters of the steam generator.

7. Conclusions

Fig. 6 Actual and simulated results with linear sigmoid-sigmoid function.

From the study of the boiler operation and artificial neural networks, following conclusions can be drawn: (1) Neural networks have proved to be very efficient with the data sets having non linear relationships.


(2) The ‘Neuro-Solutions’ software worked effectively for the prediction of steam flow rate and steam pressure in the boiler. The dynamic behavior of the physical variables through the proposed ANN model is satisfactory and consistent with the practical experience. (3) The change in learning rate and momentum term increases the rate of convergence significantly. Also this ANN model, unlike the mathematical model can be applied to the entire boiler, i.e., separate ANN model is not required for the different parts of the boiler.

References [1]

[2]

X.J. Liu, L. Felipe, C.W. Chan, Neurofuzzy network modeling and control of steam pressure in 300 mw steam boiler system, Engineering Applications of Artificial Intelligence 16 (2003) 431-440. M.G. Allen, C.T. Butler, S.A. Johnson, E.Y. Lo, F. Russo, An imaging neural network combustion control system for utility boiler applications, Combustion and Flame 94 (1993) 205-214.

[3]

15

P.C. Chattopadhyay, Boiler Operation Engineering, 2e Tata Mc-Graw Hill Publication, New Delhi, 2002. [4] E.J. Adam, J.L. Marchetti, Dynamic simulation of large boilers with natural recirculation, Computers and Chemical Engineering 23 (1999) 1031-1040. [5] S.L. Tarasiewicz, P. Radziszewski, Oil fired boiler simulation, Mathl. Comput. Modelling 14 (1990) 1075-1078. [6] A. Lawa, C. Maffezzoni, G. Benelli, Validation of drum boiler model through complete dynamic tests, Control Engineering Practice 7 (1999) 11-26. [7] I.M. Alatiqi, A.M. Meziou, Simulation and parameter scheduling operation of waste heat steam boilers, Computers Chemical Engg. 16 (1992) 51-59. [8] R. Masini, E. Padovani, M.E. Ricotti, Dynamic simulation of steam generator by neural networks, Nuclear Engineering and Design 187 (1999) 197-213. [9] S. Haykin, Neural Networks: A Comprehensive Foundation, Pearson Education (Singapore) Ltd., 2001. [10] J.M. Zurada, Introduction to Artificial Neural Networks, Jaico Publishing House, Mumbai, 1996. [11] J.K. Wu, Neural Networks and Simulation Methods, Marcel Dekker Inc., New York, 1994. [12] K. Mehotra, C. Mohan, S. Ranka, Elements of Artificial Neural Networks, The MIT Press, Cambridge, 1997.


Design and Development of a Laboratory Scale Biomass Gasifier S.J. Ojolo1 and J.I. Orisaleye2 1. Mechanical Engineering Department, University of Lagos, Lagos 101017 Nigeria 2. Mechanical Engineering Department, Lagos State University, Lagos 101017 Nigeria

Received: February 10, 2010 / Accepted: March 24, 2010 / Published: August 31, 2010. Abstract: A laboratory scale downdraft biomass gasifier was designed to deliver a mechanical power of 4 kW and thermal power of about 15 kW. The gasifier was manufactured as a single piece having a water seal and cover. The gasifier was tested in natural downdraft and forced downdraft mode. Ignition of the fuel beneath the grate, during natural downdraft mode, using wood shavings as fuel, produced gas which burned with a blue flame for 15 minutes. Ignition at the throat, using either palm kernel shells or wood shavings, during the natural downdraft mode, the gasifier did not produce syngas. During the forced downdraft mode, fuel was ignited at the throat. Gasification was successful with the palm kernel shells, during forced downdraft, which produced gas which burned steadily with luminous flame for 15 minutes per kilogram of biomass fed. However, wood shavings experienced some bridging problems during the forced downdraft mode of operation. The fuel conversion rate of the gasifier, when using palm kernel shells as fuel in forced downdraft mode, was 4 kg/h. Forced downdraft mode of operation yielded better results and is the preferred operation of the gasifier. Key words: Biomass, gasifier, design, downdraft, energy.

1. Introduction Agriculture and energy have always been tied by close links, but the nature and strength of the relationship have changed over time [1, 2]. The linkages between agriculture and energy output markets weakened in the twentieth century as fossil fuels gained prominence in the transport sector. The use of renewable resources would contribute to a country’s economic growth, especially in developing countries, many of which have abundant biomass and agricultural resources that provide the potential for achieving self-sufficiency in materials [3]. In most African countries, biomass continues to be the main energy source for subsistence activities such as cooking, heating and lighting. Solid biofuels such as fuel wood, charcoal and animal dung constitute by far the largest Corresponding author: S.J. Ojolo, senior lecturer, research fields: design and manufacturing, renewable energy. E-mail:[email protected].

segment of the bioenergy sector, representing a full 99 percent of biofuels [4, 5]. Gasification means the transformation of solid fuels into combustible gases in presence of an oxygen carrier (air, O2, H2O, CO2) at high temperatures. It is a process for converting carbonaceous materials to a combustible or synthetic gas like biomethane or producer gas [6]. Biomethane can be used like any other fuel, such as natural gas, which is not renewable [7]. The gasification process occurs at temperatures between 600-1,000 degrees Celsius and decomposes the complex hydrocarbons of wood [8]. The gasification process, with high temperature, produces ash and char, tars, methane, charcoal and other hydrocarbons. The corrosive ash elements such as chloride and potassium are removed, allowing clean gas production from otherwise problematic fuels [9]. Conversion of solid biomass into combustible gas has all the advantages associated with using gaseous and liquid fuels. Such

Design and Development of a Laboratory Scale Biomass Gasifier

advantages include clean combustion, compact burning equipment, high thermal efficiency and a good degree of control. Biomass is also economic in places where biomass is already available at reasonable low prices or industries using fuel wood [6]. Biomass gasifiers are reactors that heat biomass to produce a fuel gas that contains from one-fifth to one half (depending on the process conditions) the heat content of natural gas. A biomass gasifier converts solid fuel such as wood waste, saw-dust briquettes and agro-residues converted into briquettes into a gaseous fuel through a thermo-chemical process and the resultant gas can be used for heat and power generation applications [10, 11]. Biomass gasifiers have been classified based on their operation principles such as gasification and product temperature, oxygen requirements, product gas composition amongst others. The major types of gasifiers are updraught or counter current gasifier, downdraft or co-current gasifiers, cross-draft gasifier and fluidized bed gasifier [12]. The throated downdraft gasifier is suitable for biomass gasification, has a low tar yield, high carbon conversion, low ash carry over and simple construction and operation. However, it has a high gas exit temperature, requires uniformly sized feed stock and limited moisture content of feed. This work presents the design of a laboratory scale biomass downdraft gasifier.

2. Theory of Gasification

17

–242 MJ/kmol The pyrolysis reactions involve the cracking of the heavier biomass molecules into lighter organic molecules and carbon monoxide:

The reduction/gasification reactions involve, mainly, the gasification of tar, depending on the technology used. They include: The Boudouard reaction: The water gas reactions:

Methane synthesis reactions:

3. Design of the Gasifier/Reactor A laboratory scale biomass gasifier is for a micro scale application which is to produce mechanical power of about 1 to 7 kW. A mechanical power of 4 kW is assumed and the design of the gasifier is based on this. The design of the reactor is basically empirical, that is, implied from charts based on past experiences. 3.1 Power Consumption of the Gasifier For an engine with a compression ratio of 9.5:1, the effeciency has been estimated to be 28 per cent [12]. Therefore, the thermal power in the gas can be

The substance of a solid fuel is usually composed of the elements carbon, hydrogen and oxygen. In addition, there may be nitrogen and sulphur present in small quantities. Biomass gasification in air can be expressed in three stages which are oxidation or combustion, pyrolysis and reduction or gasification [12]. Combustion/oxidation reactions provide the heat energy required to drive the pyrolysis and char gasification reactions: –111 MJ/kmol –283 MJ/kmol

estimated as

If the thermal effeciency of the gasifier is taken at 70 per cent, the thermal power consumption at full load can be estimated as


18

3.2 Biomass Consumption of the Gasifier A heating value of biomass with 14% moisture content is taken to be 17,000 kJ/kg, according to Ref. [12]. The biomass consumption of a gasifier can be estimated as [12, 13].

3.3 The Hearth Load: Specific Gasification Rate and Specific Solid Flow Rate The hearth load, Bg, is defined as the amount of producer gas reduced to normal (p, T) conditions, divided by the surface area of the throat at the smallest circumference and is expressed in m3/(cm2xh) [12].

Sivakumar et al. [17] discovered from their model that for throat angles of about 45°, the cumulative conversion effeciency is increased while larger angles of about 90° decrease the cumulative conversion effeciency because of a decreased temperature for larger throat angles due to the divergent effect and the reaction rate. Venselaar [18] also recommended, after comparison of the design characteristics of a number of gasifiers, that the throat inclination should be around 45° to 60°. A throat angle of 60° is used. 3.5 Sizing of the Fire Box or Hearth Diameter of the fire box or hearth, dh is a function of throat diameter and can be estimated from Fig. 1a using

This may be referred to as the specific gasification rate (SGR), which is the volumetric flow rate of gas per unit area based on throat diameter, the gas volume being measured at the standard conditions [14, 15]. The hearth load can also be expressed as the amount of dry fuel consumed divided by the surface area of the narrowest constriction, Bs, and is expressed in kg/(cm2xh) [12]. This may also be referred to as the specific solid flow rate (SSR) which is the mass flow of fuel measured at throat [14, 15]. One kilogramme of

3.6 Nozzle Design and Air Blast Velocity The height of nozzle plane above the smallest cross section of the throat is a function of the throat diameter and can be evaluated from Fig. 1b, The ratio between the nozzle flow area and the throat area is a function of the throat diameter and is given from Fig. 1c as

dry fuel under normal circumstances produces about 2.5 m3 of producer gas [12, 14, 16]. Thus [12, 14]

The recommended value for Bg falls in the range of

Where, An is the total nozzle area. It is assumed that the gasifier will be equipped with 5 nozzles as recommended by Shrinivasa and Mukunda [19] for operating slow two-cycle engines. Hence,

0.30 to 1.0 [12, 14, 16]. Taking the value of Bg as 0.3, Sivakumar et al. [14] suggested optimum results are obtained when the angle of inclination of the nozzles is 3.4 Throat Sizing The cross sectional area of the throat is thus

The diameter of throat, dt, can be calculated using

between 10° and 25°. An inclination of 15° is used. The nozzle tip ring diameter dnt is also a function of the throat diameter as seen in Fig. 1d. The ratio between the nozzle tip ring diameter and the throat diameter is


a.

c.

19

b.

d.

Fig. 1 a. Diameter of the fire box, dr, as a function of the throat diameter, dt ; b. Height of the nozzle plane above the throat, hnt, as a function of the throat diameter; c. Ratio between nozzle flow area, An, and throat area, At, as a function of the throat diameter; d. Nozzle tip ring diameter, dnt, as a function of the throat diameter, dt [1].

3.7 Air Inlet and Outlet The air blast velocity (Vb) can be estimated by equating the volumetric flow rate of the producer gas through the throat to the volumetric flow rate of air through the nozzle. The volumetric flow rate of producer gas through the throat is estimated using

Using this flow rate as the flow of air through the nozzle,

The general range for air inlet velocity is 6 m/s to 10 m/s [17]. The dimensions for the air inlet can be obtained using the continuity equation. By taking the velocity of air to be 6 m/s,

For

a

circular

opening,

the

diameter

is

. The gas inlet is taken to be 25 mm. The gas outlet is taken to be 20 mm. 3.8 Length of Reduction Zone Sivakumar et al. [17] proposed that for a throat diameter of about 100 mm and for a throat angle of between 45 and 90 degrees, the reduction zone with a


20

length above 150 mm gives an optimum cumulative conversion effeciency. However, SERI [20] proposes that the height of the reduction zone should equal the diameter of the throat. The reduction zone is designed with a length of 80 mm. 3.9 Height of the Hoper The height of the hoper is decided on the basis of the feedstock which it will be required to hold within the period of operation. The period of operation is taken to be 2 hours since it is laboratory scale. Therefore, given the biomass consumption rate as 4.32 kg/h, the total biomass consumption estimated to be consumed in 2 hours is 8.64 kg. The bulk density of wood is between 300 and 550 kg/m3 depending on the moisture content. Assuming the value of the bulk density is taken to be 500 kg/m3, the total volume of the hoper is estimated as:

The height of a cylindrical reactor is

Fig. 3 The laboratory scale biomass gasifier.

A height of 400 mm is taken. The biomass gasifier is shown in Fig. 2. Fig. 3 shows the picture of the gasifier.

4. Preliminary Tests Carried Out on the Biomass Fuels Wood shavings and palm kernel shells were used as the biomass fuel/feedstock for testing the performance of the manufactured biomass gasifier. Preliminary experiments were carried out on the biomass fuels to determine some of their properties which are critical to the operation of the gasifier. The properties that were determined include the moisture content, bulk and apparent densities and bed voidage. 4.1 Determination of Moisture Content of Biomass The moisture content determined on dry weight basis (MCD) and wet basis (MCW) were estimated using Fig. 2 Orthographic views of the laboratory scale biomass gasifier.


4.2 Determination of Bulk Density The bulk densities of the wood shavings and the palm kernel shells were determined using 4. 3 Determination of Apparent Particle Density The apparent particle density is the density of a biomass particle with the pore inherently present in it. The apparent densities were determined using

21

draft mode, a blower with power rating of 3.5 kW was attached to the gasifier’s inlet to introduce air into the air jacket which delivers it to the nozzles. The fuel was fed to the throat level and was ignited at the throat while the blower was operating. After sufficient temperature was reached, the fuel was loaded in multiples of 1 kg. The gas produced was ignited and the characteristic of the gas and flame were observed. The consumption rate of biomass was also observed. The experiment was carried out for both the wood shavings and palm kernel shells. The fuel conversion rate was estimated using

4.4 Determination of Bed Voidage The bed voidage is the ratio of the inter-particle void space to the total volume. The bed voidage was determined using [21] 4.5 Tests Carried Out on the Gasifier The performance of the laboratory scale biomass gasifier was tested using the biomass fuels and in natural- and forced-draft modes. 4.5.1 Natural Downdraft Mode During operation of the gasifier in the natural convection mode, the fuel wood shavings was fed into the gasifier until it reached above the throat level. It was ignited from the bottom of the grate using a torch. Air was allowed to flow through the air inlet by natural convection. The gas produced was ignited. The properties of the gas and the flame were observed. In another experiment, the fuel was fed to the hearth level and the ignition was carried out by introducing a flame in the throat. The operation of the gasifier was also observed. The second experiment was repeated for the palm kernel shells. 4.5.2 Forced Downdraft Mode During operation of the gasifier in the forced down-

5. Results and Discussions 5.1 Determination of Biomass Properties The results for the determination of the properties of the biomass fuels are given in Table 1. 5.2 Results of Experiments Carried Out on the Gasifier Operation in Natural Downdraft Mode The operation of the gasifier with wood shavings in the natural downdraft mode took about 35 minutes for the initial start up and production of combustible gas when ignited from under the grate. The gas burned with a blue flame similar to the butane gas for a short period of about 10 minutes before the production of the syngas was stopped. The pressure at which the gas exited the outlet was low. It was observed further that the heat remained in the reduction zone and thus the oxidation zone at the throat did not combust the biomass wood Table 1 Properties of the biomass fuels. Property Moisture content (dry weight basis) (%) Moisture content (wet basis) (%) Bulk density (g/cm3) Apparent particle density (g/cm3) Bed voidage

Palm shell

kernel Wood shavings

9.8

19.3

8.9 0.5 0.82 0.39

16.2 0.05 0.3 0.83

22


Table 2 Results from the operation of the gasifier using palm kernel shells. Start up time Mass of biomass consumed Time taken to operate steadily Fuel conversion rate

10 minutes 2 kg 30 min 4 kg/h

shavings in the zone. There was, in addition, no observable decrease in the bed height during the process. Syngas was not produced when wood shavings were ignited at the throat. Instead, the biomass only burned but did not gasify. A lot of smoke was also produced before the wood shavings began to burn. Palm kernel shells also burned but did not gasify when operated in the natural draft mode. It was noticed that the ignition from the bottom of the grate was not an efficient method of ignition. Also, the gasifier could not operate by natural draft and it needed a source of forcing air through the bed. 5.3 Operation in Forced Downdraft Mode The operation of the gasifier in forced downdraft mode using wood shavings as fuel produced combustible gases, but a bridging was noticed which was caused by the char produced from the wood shavings which did not allow the gas produced to flow through the bed to the outlet. The bridging problem has been observed for fluffy or loose biomass by Kumararaja [22] for gasification of groundnut shells and Rudakova [23] for sawdust. In addition to the bridging problem, it was observed that there was no free flow of biomass within the gasifier into the throated region. The possible cause for the hindered flow is the inherent properties of the biomass as observed also by Kumararaja [22]. The operation of the gasifier in forced downdraft mode using palm kernel shells as fuel produced the results in Table 2. During the gasification of the palm kernel shells, the start up time was about 10 minutes. A lot of smoke was produced during the start-up after which combustible gases were produced steadily. The gas produced, when ignited, burned a luminous flame as that obtained

during the flaring of natural gas. Unlike the wood shavings, the palm kernel shells flowed freely. There were no bridging problems observed. It was also observed that a lot of smoke and tar oil was produced initially when the gasifier is loaded with fuel.

6. Conclusions A laboratory scale downdraft biomass gasifier was designed to deliver a mechanical power of 4 kW and thermal power of about 15 kW. The design was largely empirical, that is, based on past experience. The biomass gasifier was manufactured as a single piece having a water seal and cover. The laboratory scale biomass gasifier has a capacity of holding 8.64 kg of feedstock. The hearth and throat diameter are 238 mm and 68 mm respectively. It had five nozzles, 10 mm diameter, for the injection of air. The gasifier is lagged along its length and throat using fibre glass material. Palm kernel shells had moisture content of between 8% and 10% while wood shavings had moisture content between 16% and 20%. The bulk density of palm kernel shells is estimated to be 0.5 g/cm3. Palm kernel shells also have an apparent density estimated as 0.82 g/cm3 and with a bed voidage of 0.39. Wood shavings have a bulk density estimated to be 0.05 g/cm3. The apparent density of wood shavings is 0.3 g/cm3 and the bed voidage is 0.83. The biomass fuel conversion rate is 4 kg/h. The gasifier was suitable for gasifying palm kernel shell, but bridging problems were experienced during gasification of wood shavings. Forced downdraft mode of operation yielded better results and is the preferred operation of the gasifier.

References [1] [2] [3]

[4]

Food and Agricultural Organisation, FAO, Wood Gas as Engine Fuel, Italy, 1986. K.R. Anil, G. Yogi, Alternative Energy in Agriculture, Vol. II, CRC Press, 1986, pp. 83-102. J.J. Ramrez, J.D. Martnez, S.L. Petro, Basic design of a fluidized bed gasifier for rice husk on a pilot scale, Latin American Applied Research, ISSN 0327-0793 version impressa, Argentina, Oct. 2007. D.Md. Miah, A.R. Harun, M.Y. Shin, Wood fuel use in the


[5] [6]

[7]

[8]

[9] [10] [11]

[12]

[13]

[14]

traditional cooking stoves in the rural floodplain areas of Bangladesh, a socio-environmental perspective, Biomass and Bioenergy 33 (2009) 70-78. The Need Project, Elementary Energy Infobook, The Need Project, Manssas, 2008. P. Tobias, Solid Biomass Gasification as Fuel Source for Fuel Cells; Workshop: Hydrogen- and Fuel Cell Based Energy Systems in a Future Sustainable Energy World; Research Group for Sustainable Energy Technology, 2004. M.S. Leland, Anaerobic Digester Lagoon with Methane Gas Recovery: First Year Management and Economics, EcoGeneration Solutions, LLC available online at: www.biomassgasification.com, 2001. J. Rezaiyan, N.P. Cheremisinoff, Gasification Technologies: A Primer for Engineers and Scientists, CRC Press, USA, 2005. D. Cassidy, Biomass Gasification, available online at: www.forestencyclopedia.net on Dec. 31, 2008. Chanderpur Works, Biomass Gasifier Projects, available online at: www.chanderpur.com on Dec. 12, 2008. Global Collaborations, Global Energy Vision Project of Global Energy Collaboration, USA, available online at: www.biomassgasifier.com, 2004. Food and Agriculture Organisation of the United Nations, FAO, The State of Food and Agriculture - Biofuels Prospects Risks and Opportunities, Rome, 2008. N.S. Rathore, N.L. Panwar, Y.V. Chiplunkar, Design and techno economic evaluation of biomass gasifier for industrial thermal applications, African Journal of Environmental Science and Technology 3 (1) (2009) 6-12. S. Sivakumar, K. Pitchandi, E. Natarajan, Design and analysis of down draft gasifier using computational fluid

[15]

[16]

[17]

[18] [19]

[20]

[21]

[22]

[23]

23

dynamics, Department of Mechanical Engineering, College of Engineering, Anna University, Guindy, India, 2006. A. Jain, Design Parameters for a Rice Husk Throatless Gasifier, Agricultural Engineering International: the CIGR Ejournal, Manuscript EE 05 012. 8, 2006. T.B. Reed, A. Das, Handbook of Biomass Downdraft Gasifier Engine Systems, Solar Energy Research Institute, USA, 1988. S. Sivakumar, K. Pitchandi, E. Natarajan, Modelling and simulation of down draft wood gasifier, Journal of Applied Sciences 8 (1) (2008) 271-279. J. Venselaar, Design rules for down-draught gasifiers, a short review, IT Bandung, Indonesia, 1982. U. Shrinivasa, H.S. Mukunda, Wood gas generators for small power (~5hp) requirements, Sadhana, India, J. of Current Science 7 (2) (1984) 137-154. SERI, Generator gas-the Swedish experience from 1939-45 (translation), Solar Energy Research Institute (1536 Cole Boule Varch, Golden Colorado 80401, USA), reproduced by US Department of Commerce, NTIS, SERI, 1979, SP-33-140. R.W. Missen, C.A. Mims, B.A. Saville, Introduction to Chemical Reaction Engineering and Kinetics, John Wiley & Sons, Inc., USA, 1999. L. Kumararaja, Development of gasifier suitable for nonwoody bioresidues for electric power generation, Project Report, Department of Science, Technology and Environment, Govt. of Puducherry, Puducherry-605 014, 2009. I. Rudakova, Use of biomass gasification for transport, Master’s Thesis, Faculty of Technology, Lappeeranta University of Technology, Finland, 2009.


Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator K. Oshima and Y. Uchiyama Graduated School of Systems and Information Engineering, Laboratory of Advanced Research B0726, University of Tsukuba, 1-1-1, Tennodai, Tsukuba Science City, Ibaraki 305-8573, Japan

Received: April 02, 2010 / Accepted: May 17, 2010 / Published: August 31, 2010. Abstract: The authors propose a new closed cycle oxy-fuel gas turbine power plant that utilizes a nuclear heat generator. A pressurized water reactor (PWR) is designed to supply saturated steam to an oxy-fuel gas turbine for a specific power output increase. The saturated steam from the reactor can have lower pressure and temperature than those of an existing PWR. In this study, the authors estimated plant performances from a heat balance model based on a conceptual design of a hybrid plant and calculated the generating costs of the proposed plant from the Japanese cost data of an existing PWR plant and an liquefied natural gas (LNG) combined cycle gas turbine plant. The generating efficiency of an oxy-fuel gas turbine plant without a nuclear steam generator is estimated to be less than 35%. Based on this efficiency, with a nuclear steam generator contributing to the power output of the proposed hybrid plant, the corresponding generating efficiency is estimated to be around 45%, even if the steam conditions are lower than in an existing PWR. The generating costs are 15–20% lower than those calculated from the weighted heat performances of both an oxy-fuel gas turbine plant without a nuclear steam generator and an existing PWR plant. Key words: Natural gas, nuclear energy, hybrid power plant, gas turbine, oxy-fuel combustion.

1. Introduction Progress has been made in improving the generating efficiency of fossil fuel fired power plants by designing plants with high-temperature and high-pressure working fluid conditions. Plant efficiency has also been improved by technological integration such as combining steam and gas turbines in a power plant, or using a heat and power co-generation system. Along the same lines, a new concept to improve plant performance involves combining a fossil fuel fired steam power plant with low condition steam from an external source. This steam is utilized to increase the steam turbine power of the plant. One of the suitable methods to generate this steam is by using a biomass combustion boiler [1]. This hybrid-type fossil fuel and biomass fired power

Corresponding author: K. Oshima, Ph.D., research field: power engineering. E-mail: [email protected].

plant is effective at improving plant performance compared to conventional biomass steam power generation, where the steam conditions cannot be raised due to the limitations of the combustion boiler. The steam conditions are also limited for a light water nuclear power plant, which is necessary for rigorous plant safety. The steam of a pressurized water reactor (PWR) is supplied to a steam turbine at the same level as that of a biomass combustion boiler. The authors examined the combination of a fossil fuel fired steam plant and a PWR and found that the nuclear power generating performance can equal or be greater than that of an existing PWR plant. In this paper, the authors propose a new plant design using a particular kind of internal combustion power plant fueled by liquefied natural gas (LNG)—a closed cycle oxy-fuel gas turbine power plant combined with a nuclear heat generator. This hybrid plant is expected to improve plant performance by

Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator using a nuclear reactor to contribute to the specific power output of a gas turbine. The power generating performances were evaluated from heat balance models based on a conceptual plant design. The generating costs of the proposed plant were also estimated from the Japanese cost data for an existing PWR plant and an LNG combined cycle gas turbine plant.

2. Analysis of Generating Performances In

this

chapter,

the

authors

explain

the

performances of the two plants examined for our analysis: a closed cycle oxy-fuel gas turbine plant

Table 1

Computational assumptions.

Combustor inlet pressure (MPa) Net power capacity (MWe) Turbine inlet temperature (deg C) Turbine outlet pressure (MPa) O2 excess ratio (%) Adiabatic efficiency of gas turbine (%) Adiabatic efficiency of pump (%) Generator and mechanical efficiency ηgm (%) Combustor pressure drop (%) Efficiency drop by auxiliaries (reference) dau (%) Efficiency drop by auxiliaries (proposed) d´au (%) Heat value of LNG (LHV) qLNG (MJ/kgLNG) O2 production work wO2 (MJ/kgO2) CO2 capture work wCO2 (MJ/kgCO2)

used as a reference and the proposed hybrid plant. A closed cycle oxy-fuel gas turbine plant is expected to be installed as an efficient carbon capture and storage system. The steam or hot water injected into the combustor assumes the role of the major working

(1) H2O p1: 5.00 T1: 179 h1: 0.76 m1: 144

(3) O2 p3: 5.00 (4) T3: 25 m3: 84 p4: 4.75 T4: 1300 h4: 4.77 Combustor m4: 248

(2) CH4 p2: 5.00 T2: 25 m2: 20

unit of injected H2O in a combustor is expressed as (1)

where n equals the mole number of LNG fuel (100%

2.00 - 5.00 200 - 1000 1300 0.008 5.0 92 78 98 5.0 2.0 2.7 50 1.47 0.73

pi: MPa Ti: deg C hi: MJ/kg mi: kg/s

Turbine 344 MWe G Generator (5) p5: 0.008 T5: 263 h5: 2.35

fluid for this plant. The ideal chemical reaction per follows: nCH 4 + 2nO 2 + H 2 O → nCO 2 + (2n + 1)H 2 O

25

Recuperator

(6) p6: 0.008 T6: 40 h6: 1.97 CO2

CH4) injected into the combustor of the gas turbine.

(8) p8: 5.00 T8: 25 h8: 0.11

The mole number n depends on the injection H2O temperature; a higher temperature resulting in a

Pump

(7) p7: 0.008 T7: 25 h7: 0.10

Condenser H2O

smaller number for the purpose of attaining the same

Fig. 1

turbine inlet temperature. The turbine power capacity

the fuel, oxygen, and saturated steam of a nuclear generator is equal to the saturated steam pressure of a conventional nuclear reactor, which is about 5.00 – 6.00 MPa. The net power capacities of the proposed plants ranged from 200 to 1000 MWe based on the capacities of existing power plants. Other assumptions such as the loss ratios and efficiencies of facilities were referenced from preceding studies [4, 5]. The schematic diagram of a reference plant is shown in Fig. 1 for the case of a combustor inlet pressure of 5.00 MPa and a net power capacity of 344 MWe. A heat balance model can be constructed using the thermodynamic properties of the working fluids at

per unit of LNG fuel will increase with a higher H2O temperature. There are some previous studies on hybrid plants where the injection H2O was heated by renewable energy such as solar thermal energy [2], and waste heat energy [3]. In this study, the authors propose to utilize a PWR as a heat generator for the injected H2O. The computational assumptions to construct a model for the hybrid plant are based on the plant performance of an oxy-fuel gas turbine, as shown in Table 1. The performance of a power generating plant was analyzed in a gas turbine combustor inlet pressure range of 2.00-5.00 MPa. The pressure supplied from

Schematic diagram of reference plant.

26

Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator

process points (1)–(8) in Fig. 1 based on the pressure assumptions shown in Table 1. The net power capacity, W (MWe), of a reference plant is defined as follows: W　 = ηgm (1 − d au ){m4 (h4 − h5 ) − m1 (h8 − h7 )} (2) 　　　　− m3 wO2 − 2.75m2 wCO2

Table 2 Major heat balance parameters of reference plant (for the case of Fig. 1). W (MWe) 344

The net power capacity of W´ is generated by input energy Q´, which is supplied by both LNG fuel

Η (%) 34.5

(3) (2) O2 CH4 (1) p2: 5.00 p3: 5.00 H2O T2: 25 T3: 25 p1: 5.00 m2: 20 m3: 84 T1: 264 h1: 2.79 Combustor m1: 256

W is a value obtained by subtracting the auxiliary power, O2 production work, and CO2 capture work from the gross gas turbine power output. The amount of heat generated by the combustion of LNG fuel Q (MWt) is calculated by: Q = m2 qLNG (3) where Q is calculated by multiplying the fuel mass flow rate (kg/s) by the lower heating value of LNG (MJ/kgLNG). Eventually, the net power generating efficiency, η (%), of the reference plant is expressed by the ratio of Q to W, as shown in Eq. (4). W η= (4) Q The results of the major heat balance parameters of the reference plant are summarized in Table 2. The net efficiency can also be estimated for other cases, with combustor inlet pressures between 2.00-5.00 MPa, resulting in 32.6 % for 2.00 MPa and 34.5 % for 5.00 MPa. Next, the authors describe the performance of the proposed hybrid plant. A PWR secondary coolant system is introduced to the closed-cycle oxy-fuel gas turbine, as shown in Fig. 2. The thermodynamic properties of the working fluids are also determined in series to construct heat balance models, while the process points increased by 1, resulting in points (1)–(9). A PWR reactor is added to the schematic diagram of the reference oxy-fuel gas turbine plant. The net power capacity, W´ (MWe), of the whole proposed plant is represented in Eq. (5). W ′ = η gm (1 − d au' ){m4 (h4 − h5 ) − m1 (h8 − h7 )} (5) 　　　　− m3 wO2 − 2.75m2 wCO2

Q (MWt) 997

(4) p4: 4.75 T4: 1300 h4: 4.77 m4: 360

pi: MPa Ti: deg C hi: MJ/kg mi: kg/s

Turbine 600 MWe G Generator (5) p5: 0.008 T5: 257 h5: 2.54

(9) p9: 5.00 T9: 153 h9: 0.65

(6) p6: 0.008 T6: 40 h6: 2.16 Recuperator CO2 (8) p8: 5.00 T8: 25 h8: 0.11

Fig. 2

Pump

(7) p7: 0.008 T7: 25 h7: 0.10

Condenser H2O

Schematic diagram of proposed plant.

combustion and a nuclear fission reaction. The net power capacity contributed just by the LNG system is defined as W´LNG (MWe) in Eq. (6) using the net thermal efficiency, η´LNG , of the reference plant. ′ = QLNG ′ η LNG ′ WLNG (6) In order to derive the net power capacity contributed by nuclear energy, η´LNG is assumed to be equivalent to η of Eq. (4). Then, the net power capacity contributed just by the nuclear energy of a PWR, W´PWR (MWe), can be obtained by Eq. (7). ′ ′ = W ′ − WLNG WPWR

(7)

The inlet heat of the whole plant is expressed as the sum of the two different heat energies supplied by the gas turbine combustor and nuclear reactor. Q´ (MWt) in Eq. (8) equals to the sum of Q´LNG (MWt) in Eq. (9) generated by the heat of LNG fuel combustion and Q´PWR (MWt) in Eq. (10) generated by the fission reaction of the PWR. ′ + QPWR ′ Q′ = QLNG ′ = m2 qLNG QLNG ′ = m1 (h1 − h9 ) QPWR

(8) (9) (10)

The net generating efficiency of the whole plant, η´, and corresponding net generating efficiency contributed


′ WPWR (MWe)

′ QPWR (MWt)

′ η PWR (%)

256

550

46.6

(b) LNG system and the whole plant

′ WLNG

′ QLNG

′ η LNG

W′

Q′

η′

(MWe) 344

(MWt) 997

(%) 34.5

(MWe) 600

(MWt) 1547

(%) 38.8

by the PWR system, η´PWR , are shown in Eq. (11) and Eq. (12).

W′ Q′ W′ = PWR ′ QPWR

η′ = ′ η PWR

(11)

50.0 Generating efficiency [%]

Table 3 Major heat balance parameters of a hybrid plant (for the case of Fig. 2). (a) PWR system

27

45.0 40.0 35.0 30.0 25.0

Saturated steam pressure [MPa] 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Proposed 44.8 45.3 45.7 46.0 46.2 46.4 46.6 Existing 28.6 29.5 30.3 30.9 31.4 31.8 32.2

Fig. 3 Comparison of PWR generation efficiencies (“existing” estimated from rankine cycle model).

3.1 Cost Estimation of Generating Plant (12)

From both Eq. (11) and Eq. (12), the major heat balance parameters of the hybrid plant are calculated as shown in Table 3. From the PWR system contribution shown in Table 3(a), the hybrid plant can attain over 45% of the corresponding generating efficiency, which is higher than the 33% of the conventional PWR power plant even if a similar steam condition of 5.00 MPa saturated steam is applied to both plants. The generating efficiency of a hybrid plant is affected by the inlet pressure of the gas turbine. In Fig. 3, the net corresponding generating efficiency contributed by the proposed PWR system is plotted in an inlet pressure range of 2.00-5.00 MPa and is compared with the net generating efficiency of an existing PWR power plant. As noted from Fig. 3, the net generating efficiency of the proposed nuclear energy system is estimated to be between 44.8-46.6%, which is approximately 10% points higher than that of an existing PWR plant.

3. Economic Analysis of Power Generation The generating costs of the reference plant and

The generating costs of the power plants are quoted from Japanese model plant data for an existing LNG combined cycle power plant and a PWR power plant estimated in preceding studies [6-8]. As shown in Eq. (13), the generating cost consists mainly of variable cost and fixed cost, where the former represents fuel cost and the latter includes both investment cost and operation and maintenance (O&M) cost. LNG fuel and nuclear fuel are consumed in the hybrid plant. The variable cost can be calculated by the two different fuels consumptions obtained in the previous chapter. The fuel costs are calculated using the available data of 0.158 ¥/MJt for nuclear fuel and 0.514 ¥/MJt for LNG fuel [7]. The fixed cost can be calculated using the capital cost, annual operating hours at the rated capacity, and annual carrying charge rate, as shown in Eq. (13). The annual carrying charge rate and operating hours are taken from preceding studies [7]. 3.6　 × FP　 CC　 × CCR GC = VC + FC = 　 + (13) GE OH GC: Generating Cost (¥/kWhe); VC: Variable Cost (¥/kWhe); FC: Fixed Cost (¥/kWhe); FP: Fuel Price (¥/MJt); GE: Generating Efficiency (net) (%); CC: Capital Cost (¥/kWe); CCR: Carrying Charge Rate (Annual) (%); OH: Operating Hours (Annual) (h).

proposed hybrid plant are next estimated for different net power capacities from 200 to 1000 MWe with a gas turbine combustor inlet pressure of 5.00 MPa.

3.2 Capital Requirements The capital cost is derived from Eq. (14). It is given

28


by the capital requirement, which is the sum of the direct capital requirement, indirect capital requirement, and contingency, divided by the power capacity. CC =

CR DCR + ICR + Co = PC PC

(14)

CC: Capital Cost (¥/kWe); CR: Capital Requirement (¥); DCR: Direct Capital Requirement (¥); ICR: Indirect Capital Requirement (¥); Co: Contingency (¥); PC: Power Capacity (kWe). The direct capital requirement indicates the sum of the purchase costs for all of the equipment in the power plant; the indirect capital requirement represents all of the construction expenses, including the labor and building materials; and the contingency is an uncertainty cost related to plant design and site construction. The capital cost of the proposed hybrid plant is estimated by an updated capital cost based on a conventional PWR plant and an LNG combined cycle plant. The authors use these capital cost reference data. The capital cost is composed of DCR, ICR, and Co as indicated in Eq. (14) and those values are determined for every cost item constituting a power plant. The ratios of DCR to ICR and Co for the items are quoted from the results of the preceding study [8]. The value of the capital cost is affected by the rated capacity of a power plant. The scale factors of economy are determined for different values of every facility based on the cost experience investigated by the preceding study [8]. This study investigated proposed hybrid plants with power capacities ranging from 200–1,000 MWe. The input heat energy for a hybrid plant is generated by both a nuclear reactor and an LNG fuel combustor. For example, a 600 MWe hybrid plant can be divided into 256 MWe for the former and 344 MWe for the latter when the plant capacity is shared from the contribution of input heat. Before estimating the capital cost of the hybrid plant, we investigate the capital cost of a plant with a smaller capacity than the conventional power plant. Tables 4 and 5 show the capital requirement results for the corresponding capacities: 256 MWe for an

Table 4

CR for an existing PWR plant (108 \).

(256 MWe / 758MWt) Outdoor facilities Reactor building Administration and service building Auxiliary building Nuclear steam supply system NSSS associated systems Fuel handling and storage system Stack facilities Steam turbine Steam turbine auxiliaries Circulating water facilities Water treatment facilities Waste treatment facilities Electric facilities Other auxiliaries Indirect requirements Contingency Total Capital cost: 54.9×104 ¥/kWe

41.2 218.6 14.0 45.7 94.1 96.5 3.9 1.1 51.5 49.9 51.5 6.4 9.3 131.5 78.1 386.1 127.9 1407.4

existing PWR plant and 344 MWe for a reference closed cycle oxy-fuel gas turbine plant. Their capital requirements are estimated from the scale of economy, which is determined from the scale factors for the constituent equipment. The scale of economy has a great influence on the value of capital cost. The PWR plant cost is estimated to be 54.9 × 104 ¥/kWe for a smaller capacity of 256 MWe , as compared to 29.8 × 104 ¥/kWe for 1000 MWe. The capital requirements for a reference plant of 344 MWe are indicated in Table 5. These were estimated using the practical cost data for a conventional LNG combined cycle plant [8]. In Table 5, additional capital requirements are listed for O2 production and CO2 capture facilities, which were calculated separately based on a study of oxy-fuel gas turbines [9]. Their costs account for one-thirds of the total capital requirements. As for the estimation of the capital requirements for the plant, the cost data for the steam turbine and its balance of plant (BOP) facilities must be excluded from the reference data of an LNG combined cycle plant. The capital requirements for a 600 MWe hybrid plant

Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator Table 5

CR for reference plant (108\).

Table 6

(344 MWe / 997 MWt) Outdoor facilities Major plant building Administration and service building Appurtenant building Gas turbine and HRSG Gas turbine and HRSG auxiliaries Fuel handling and storage system Stack facilities Circulating water facilities Water treatment facilities Waste treatment facilities Electric facilities Other auxiliaries O2 production facilities CO2 capture facilities Indirect requirements Contingency Total Capital cost: 29.0×104 ¥/kWe

31.6 19.6 3.6 2.4 100.0 7.4 88.5 1.2 61.6 10.1 35.0 45.9 11.1 232.0 163.0 94.9 90.8 998.4

are summarized in Table 6. In the hybrid plant, the LNG system consists of the same facilities as the reference plant. However, the nuclear reactor’s only function is to generate saturated steam for the gas turbine in the hybrid plant, where six items among all of the facilities of the PWR power plant indicated in Table 4 should be evaluated as constituting items in the hybrid plant. Consequently, the capital cost of the nuclear contribution is estimated to be 24.2 × 104 ¥/kWe , less

29

CR for hybrid plant (108\).

(a) PWR system (256 MWe / 550 MWt) Reactor building Auxiliary building Nuclear steam supply system NSSS associated systems Fuel handling and storage system Waste treatment facilities Indirect requirements Contingency Subtotal Capital cost: 24.2×104 ¥/kWe (b) LNG system (344 MWe / 997 MWt) Outdoor facilities Major plant building Administration and service building Appurtenant building Gas turbine and HRSG Gas turbine and HRSG auxiliaries Fuel handling and storage system Stack facilities Circulating water facilities Water treatment facilities Waste treatment facilities Electric facilities Other auxiliaries O2 production facilities CO2 capture facilities Indirect requirements Contingency Subtotal Capital cost: 37.0×104 ¥/kWe (C) Whole plant (600 MWe / 1547 MWt) Total Capital cost: 31.5×104 ¥/kWe

187.5 38.1 75.4 81.7 3.3 7.7 170.1 56.4 620.1

31.6 19.6 3.6 2.4 132.5 9.4 112.0 1.5 77.7 13.1 45.0 68.1 16.0 295.6 207.6 120.9 115.7 1272.2

1892.3

than the result of Table 4. In the case of the LNG system indicated in Table 6(b), the capital cost increases to 37.0 × 104 ¥/kWe from the 29.0 × 104 ¥/kWe shown for the plant in Table 5, in spite of the same power capacity of 344 MWe. This is due to the effect of the PWR system installation in the LNG system. The LNG system in the hybrid plant generates 600 MWe of capacity by the installation of the PWR system. The equipment capacities of the gas turbine, circulating water facilities, water treatment facilities, electric facilities, etc. are required for the 600 MWe class plant.

3.3 Results of Power Generating Costs The generating cost is calculated using the fixed cost and variable cost indicated in Eq. (13). Before estimating the generating cost of the hybrid plant, the generating cost of the reference plant is estimated for the 256 MWe of the existing PWR plant and the 344 MWe of the reference plant, as shown in Fig. 1. Table 7 indicates the generating cost results. Based on the capital costs in Table 6 and the generating costs in Table 7, the generating cost of the hybrid plant is calculated in Table 8. The generating


Table 7

20.0

PWR plant and reference plant costs.

Existing PWR plant Power capacity (MWe) 256 Thermal capacity (MWt) 758 Generating efficiency (%) 33.8 Operating hours (h/year) 6132 Capital cost (×104 ¥/kWe) 54.9 Carrying charge rate (%) 14.40 Variable cost (¥/kWhe) 1.68 Fixed cost (¥/kWhe) 12.90 14.58 Generating cost (¥/kWhe)

Reference plant 344 997 34.5 29.0 13.80 6.54 5.36 11.90

Generating Cost [¥/kWhe]

30

non-hybrid hybrid

18.0 16.0 14.0 12.0 10.0 8.0 6.0 4.0 200

400

600

800

1000

Power Capacity [MWe]

Whole plant Power capacity

(MWe)

600

Thermal capacity

(MWt)

1547

(%)

38.8

(h)

6132

Generating efficiency Operating hours Capital cost

(×104 ¥/kWe) 31.5

Carrying charge rate

(%)

14.06

Variable cost (¥/kWhe)

3.59

Fixed cost

(¥/kWhe)

7.23

Generating cost

(¥/kWhe)

10.82

PWR Sys. LNG Sys. 256 344 550 997 46.6 34.5

34.9 29.0 14.40 13.80 1.22 5.36 8.19 6.54 9.41 11.90

cost for 600 MWe is estimated to be 10.82 ¥/kWhe , which is lower than the values in Table 7. The capital cost and generating cost are highly influenced by the capacity of the hybrid plant. The generating cost of the hybrid plant was estimated for different plant capacities from 200 to 1,000 MWe. Fig. 4 shows the results of a comparison between the generating cost of the hybrid plant with the weighted generating cost calculated by a PWR plant and an oxy-fuel gas turbine plant constructed separately. As shown in Fig. 4, the generating cost of the hybrid plant is 15-20% lower than the weighted generating cost.

Fig. 4 Cost comparison between a proposed hybrid plant and a non-hybrid plant. 20.0

Generating Cost [¥/kWhe]

Table 8 Generating cost of proposed hybrid plant (for the case of Fig. 2).

Existing Proposed

18.0 16.0 14.0 12.0 10.0 8.0 6.0 4.0 200

400

600

800

1000

Power Capacity [MWe]

Fig. 5 Nuclear power generation costs in proposed hybrid plant and existing PWR plant.

The hybrid plant is composed of nuclear facilities and oxy-fuel gas turbine facilities with CO2 capture and oxygen production facilities. The generating cost for the nuclear system can be obtained by focusing on the nuclear facilities in the hybrid plant. Fig. 5 shows a comparison of the generating costs between a nuclear system and an existing PWR plant, which are estimated for different plant capacities. The generating cost for the nuclear system is 30% lower than that of an existing PWR plant due to the high thermal efficiency and low capital cost estimated from the nuclear contribution to the specific power output of the hybrid plant.

4. Conclusions In this paper,

the authors proposed a new type of

Performance and Economic Study of Oxy-fuel Gas Turbine Power Plant Utilizing Nuclear Steam Generator nuclear power generation method utilizing a closed cycle oxy-fuel gas turbine power plant. A nuclear reactor was designed to supply saturated steam for a specific power output increase in the gas turbine. The generating performance was evaluated from a heat balance model based on the conceptual plant design. In addition, the generating costs of a hybrid plant were estimated from Japanese cost data of an existing PWR power plant and a LNG combined cycle gas turbine power plant. The corresponding generating efficiency of the PWR system was estimated to be around 45%, even if the steam condition was lower than an existing PWR power plant, assuming that the generating efficiency of the whole hybrid plant equals 35% of the oxy-fuel gas turbine system with a carbon capture system. Based on the scale of economy, the generating costs were estimated for different plants with power capacities ranging from 200 to 1,000 MWe , with a gas turbine combustor inlet pressure of 5.00 MPa. The generating costs of the hybrid plant were 15–20% lower than those of the conventional oxy-fuel gas turbine plant and an existing PWR plant. The corresponding power generation cost of the nuclear reactor in the plant was estimated to be around two thirds of the existing PWR power plant.

31

References [1]

[2]

[3]

[4]

[5]

[6]

[7] [8] [9]

Kiichiro Ogawa, Y. Torikai, Integration System of Biomass and Thermal Power Plant, Japan Society of Energy and Resources 29 (3) (2008) 184-188. C.H. Gou, R.X. Cai, H. Hong, A novel hybrid oxy-fuel power cycle utilizing solar thermal energy, Energy 32 (2007) 1707-1714. P.S. Pak, Comprehensive evaluation of a CO2-capturing NOx-free repowering system with utilization of middle pressure steam in a thermal power plant, IEEJ Trans. PE 123 (7) (2003) 808-813. O. Bolland, H.M. Kvamsdal, J.C. Boden, A comparison of the efficiencies of the oxy-fuel power cycles water-cycle, granz-cycle and matiant-cycle, Carbon Dioxide Capture for Storage in Deep Geologic Formations 1 (2005) 499-512. C.H. Gou, R.X. Cai, G.Q. Zhang, An advanced zero emission power cycle with integrated low temperature thermal energy, Applied Thermal Engineering 26 (2006) 2228-2235. H. Hondo, Y. Uchiyama, Economic Analysis of Emission Control Technologies of Fossil-fired Power Plants, CRIEPI Economic Research Center, Rep. No. Y92009, 1992. Y. Uchiyama, H. Hondo, CRIEPI Socio-economic Research Center, Rep. No. Y99507, 1999. Y. Uchiyama, CRIEPI Economic Research Center, Internal Record, No. 266, 1985. P.S. Pak, Y.D. Lee, K.Y. Ahn, Evaluation of economics of a CO2-Capturing repowering system based on oxy-fuel combustion for utilizing low pressure steam, in: Proceedings of the 24th Conference on Energy, Economy, and Environment, 2008, pp. 537-540.


A Super Low Power CMOS Receiver for High Resolution Epi-retinal Prosthesis J.W. Yang, N. Tran, S. Bai, D.C. Ng, M. Halpern and E. Skafidas Department of Electrical and Electronic Engineering, The University of Melbourne, Parkville, Victoria 3010, Australia Received: April 15, 2010 / Accepted: May 20, 2010 / Published: August 31, 2010. Abstract: The authors report a super low power Medical Implant Communication Service (MICS) band receiver for a high resolution epi-retinal prosthesis (BionicEye). The frequency shift keying (FSK) receiver consumes less than 1.5 mW power with 1 V supply. It is able to achieve a maximum data rate of 400 kb/s. This paper presents the research work carried out on designing a fully-integrated sub-threshold receiver fabricated on a 65 nm complimentary metal oxide semiconductor (CMOS) chip. In order to achieve super low power consumption, more than 90% of the transistors in all analog building blocks are operated in sub-threshold region. System level issues, such as required receiver architecture and specifications are also addressed. Key words: Super low power, fully-integrated, sub-threshold, MICS band, BionicEye.

1. Introduction A retinal prosthesis is an effective device for restoring human vision. Such a device requires wireless data link to provide captured image data to stimulate ganglion cells via an electrodes array. The receiver for this purpose demands low power consumption, for its power supply is preferably acquired by wireless coupling [1]. Moreover, most power from the wireless power link will be consumed by succeeding stimulation and electrodes driving circuits. Based on previous research [2], this paper presents a receiver design intended for a high resolution epi-retinal prosthesis with 1024 electrodes (BionicEye). Such large number of electrodes makes the power margin for the data link extremely demanding and requires at least 180 kb/s data rate to be fully functional [3]. The work proposed here is suited for this application and it has many novel features as well. Firstly, super low power consumption ( 4~5Ut) in practical applications.

designed for comparison. One is an inductor-less structure based on a common-gate current amplifier; the other one is a current-reuse structure, in which a quadrature mixer is stacked on top of a conventional cascode LNA. 4.1.1 Down Converter I (Inductor-less Structure) In the inductor-less structure, a topology for wideband input matching [9] is used as the first stage of the LNA at the cost of higher noise figure. As shown in Fig. 2(a), the first gain stage is a common-gate amplifier with shunt-feedback. An additional 20 pF on-chip capacitor C1 is added at the input port to resonate with the inductive portion of its input impedance at 400 MHz. The input impedance of the down converter as a whole is mainly determined by the first LNA stage. The real part of the input impedance can be expressed as

4. Circuit Realizations (a)

4.1 Down Converter The down converter consists of a LNA (low noise amplifier) and quadrature mixers. Matching the LNA to the 50 Ω input impedance is challenging for such a low power, 400 MHz system. Low operating frequencies would necessitate large inductors, which not only exhibit low quality factors on chip, but also require large area. Two different schematics have been investigated and Table 1

Extracted I0 and n for IBM cmos10lpe technology.

Device Type Regular NFET Low Vt NFET High Vt NFET Regular PFET Low Vt PFET High Vt PFET

n 1.84 1.74 2.06 1.81 1.66 1.72

I0 (nA) 6070 6741 2475 957 1018 331

(b) Fig. 2 Down converter schematic (a) Inductor-less structure (b) Current-reuse structure.

A Super Low Power CMOS Receiver for High Resolution Epi-retinal Prosthesis

ω2 ( gm1Cout 2 − Cgs1Cout ( go1 + goc )) + gm1gm2 ( go1 + goc ) ω4Cgs12Cout 2 + ω2 ( gm12Cout 2 − 2gm1gm2Cgs1Cout ) + gm12 gm22

(5)

While the imaginary part can be written as j(−ω3Cgs1Cout 2 + ωCout gm1 ( gm2 − go1 − goc )) ω4Cgs12Cout 2 + ω2 ( gm12Cout 2 − 2gm1gm2Cgs1Cout ) + gm12 gm22

(6)

Where gm1 and gm2 are the transconductances of M1 and M2 respectively; go1 is the on-conductance of M1; goc is the output conductance of M3 (current source); Cgs1 is the gate-source capacitance of M1 and Cout is the output capacitance of this stage. Fig. 3(a) shows the S11, real part and imaginary part of the input impedance respectively, indicating that the LNA input has been matched to 50 Ω at 400 MHz. An inverter-type current-reuse amplifier is used as the second stage in order to further increase the voltage gain and suppress the noise. However, the linearity of the down converter is degraded due to the large gain of LNA. The noise factor of the first LNA stage is crucial to the overall noise figure of the down converter. Its noise factor can be minimized by increasing the transconductance of M1 and decreasing that of M2 [9]. Thus, M1 is designed to work at deep weak inversion but M2 works in the strong inversion. This way, the transconductance ratio of M1 over M2 (gm1/gm2) could be maximized since they share the same drain current. Due to sub-threshold operation, gm1 is sensitive to the drain current value. Hence, current biasing should be used (as dicpicted in Fig. 4(a)) in parctice. A pair of single-balanced Gilbert mixers with lowpass loads is used as the qudrature mixers. M6, M7 and M8 are all in deep weak inversion and saturation, providing maximum gain with minimum current. Current source M9 is used here (known as current-bleeding technique) to carry a large part of the drain current of M7. Therefore, M5 and M6 could switch more abruptly with smaller currents and the load resistors can be made bigger. As a result, the noise figure falls and the conversion gain rises compared to the work in Ref. [2]. This down converter, as a whole, provides 30 dB voltage gain while achieving a noise

35

(a)

(b) Fig. 3 S11 and input impedance of the down converter, (a) Inductor-less structure, (b) Current-reuse structure.

(a) (b) Fig. 4 (a) Current biasing for LNA in Fig. 2(a); (b) Current biasing for LNA in Fig. 2(b).

figure of 9 dB. 4.1.2 Down Converter II (Current-reuse Structure) In the current-reuse structure, a quadrature mixer stacks on top of the cascode LNA to share the bias current [10]. The stacking is possible owing to sub-threshold operation of all transistors, and hence a low minimum saturation voltage required (around 50 mV). Fig. 2(b) shows the full schematic. The radio frequency (RF) signal at the LNA output (at the drain of M2) is AC coupled to the gate of M3. Lp and Cp form a LC (inductor-capacitor) tank resonating at 400 MHz,


which helps to filter out unwanted frequencies and improve image rejection as well. The bypass capacitor Cb must be large enough to avoid current coupling to the source of M3. In other words, the source of M3 needs to be AC ground. Thus, a properly sized ncap (NMOS in n-well) capacitor is adopted here for saving chip area. M8 works as the current-bleeding source, drawing current from M3 and extending voltage headrooms for the qudrature mixers. Input impedance matching is achieved according to Eqs. (7) and (8): 1 gm (7) Z in ( s ) = s ( L g + L s ) + + Ls sC gs ' C gs ' (8) C gs ' = C gs + C ex Where Lg and Ls are the gate and source (of M1) inductors respectively. Cgs is the inherent gate-source capacitance of M1. Together with the shunt capacitor Cex, now the equivalent gate-source capacitance becomes Cgs’, lowering the input impedance to 50 Ω. The matching parameters are shown in Fig. 4(b). All the transistors in this downcovertor are configured to work in deep weak inversion region. The shared drain current is set to be 500 μA. Since this current is critical to the whole voltage conversion gain and the noise figure, a cascode current mirror could be used in practice to bias it (as illustrated in Fig. 3(b)). This current biasing technique also removes the process and temperature variations. This down converter provides an overall voltage gain of 35 dB with a noise figure of 7.5 dB. 4.2 PLL and VCO Fig. 5 gives the architecture of the PLL. The reference frequency comes from a 300 MHz MEMS resonator (divided by 1000). Fig. 6 shows the simulated amplitude response of this MEMS nickel disk. Vibrating at first radial contour mode [11] with a maximum displacement of 2.2 nm, the disk has a quality factor of about 9,000 and its output current amplitude is about 825 nA. On-chip differential amplifiers are designed to convert the AC output current of the disk to the desired output voltage.

Fig. 5

PLL architecture.

1.0E-02 1.0E-03 Amplitude (um)

36

1.0E-04 1.0E-05 1.0E-06 1.0E-07 0

100

200 300 400 Freqency (MHz)

500

600

Fig. 6 Simulated amplitude response of the nickel disk resonator.

Fig. 7 Quadrature VCO.

A current-reuse topology that uses nMOS–pMOS cross coupled pairs is chosen for the core voltage controlled oscillator (VCO) design to save power, as shown in Fig. 7. The inductor value of the LC tank (400 MHz) is made as large as possible for larger oscillation amplitude and lower phase noise. Two 25 nH inductors are made in series. The simulated quality factor for the cascade inductors is 6.7. A pair of nCap varactors (NFET in N-well) is configured to give a VCO gain of 12 MHz/V. To calibrate the process and temperature variations, a 3-bits controlled capacitor bank is connected to the tank [12]. The VCO single output swing is transformed to a square wave through two inverters in the PLL loop


37

and its differential output is split to quadrature signals via a passive poly-phase filter. 4.3 IF Filters The transconductor using double CMOS pair is proposed in the sub-threshold IF bandpass filter design. This type of transconductor has high power supply rejection ratio (PSRR) and good linearity even when working at weak inversion [13]. Using triple well

Fig. 8 Sub-threshold double CMOS pair is used to form Nauta’s OTA.

devices for M1 and M3 can further improve its linearity. With

the

gate

of

M4

tied

to

ground,

the

transconductance can be controlled by varying the bias voltage of M1, as shown in Fig. 8. A fully-balanced transconductor cell can be formed using Nauta’s architecture [14]. A 7th-order complex bandpass filter is constructed based on a lowpass prototype (Butterworth + Notch), providing reasonable stop-band attenuation and group delay variation (50

400

13.4 -32.3

>50

400

Table 3 Summary of power consumptions. Building blocks Power consumption (μW) Down converter I LNA + I/Q mixers = 370 +220 =590 Down converter II 500 Complex IF BPF 500 Real IF BPF 95 IF VGA 30 BFSK demodulator 12 VCO 170 PLL 160 Total (I/Q channel with 1492 down converter I) Total (I/Q channel with 1392 down converter II) Total (I channel only 1057 with down converter I) Total (I channel only 967 with down converter II)

I), and its linearity (IP3, obtained when IF VGA is at minimum gain) is displayed in Fig. 13. Tables 2 and 3 summarize the performance and power consumption of the receiver in all combinations, which shows both down converters are qualified for this receiver since they have comparable performances. However, inductor-less down converter is considered to be a better choice, for it occupies a much smaller chip area. The total power consumption of the receiver is proved to be less than 1.5 mW. Therefore, it could be labeled as super low power.

Fig. 12 Voltage gain and noise figure of the receiver.

6. Conclusions

Fig. 13 IP3 plot of the receiver.

A super low power MICS band receiver was designed. Combined with electrodes stimulating and driving circuitries, the receiver will be fabricated onto a 5×5 mm2 epi-retinal prosthesis chip. With extremely low power consumption and MICS band operation, this receiver is very suitable for BionicEye and other biomedical purposes.


References [1]

[2]

[3]

[4] [5]

[6]

[7]

[8]

D.C. Ng, S. Bai, G. Felic, E. Skafidas, Closed-loop inductive link for wireless powering of a high density electrode array retinal prosthesis, in: IEEE EMCSA 2009, Adelaide, 16-19 Sep., 2009, pp. 92-97. J. Yang, N. Tran, S. Bai, D.C. Ng, M. Halpern, E. Skafidas, et al., A super low power MICS band receiver in 65 nm CMOS for high resolution epi-retinal prosthesis, in: IEEE ASICON 2009, 20-23 Oct., 2009, pp. 435-438. N. Tran, J. Yang, S. Bai, D. Ng, M. Halpern, D.B. Grayden, E. Skafidas, I. Mareels, A fully flexible stimulator using 65 nm cmos process for 1024-electrode epi-retinal prosthesis, in: IEEE EMBC 2009, 2-6 Sep., 2009, pp. 1643-1646. FCC Rules and Regulations, Table of Frequency Allocations, Part 2.106, Nov. 2002. C. Gabriel, S. Gabriel, Compilation of the dielectric properties of body tissues at RF and microwave frequencies, Final Technical Report, Occupational and Environmental Health Directorate Radiofrequency Radiation Division, Brooks Air force Base, Texas, 1996. L.H. Jung, P. Preston, G.J. Suaning, N.H. Lovell, A wideband frequency-shift keying demodulation technique for inductively powered biomedical implants, J. Australas. Phys. Eng. Sci. Med. 30 (2) (2007) 141-146. N.M. Pletcher, Micro power radio frequency oscillator design, Research Thesis, EECS Department, University of California, Berkeley, Dec. 21, 2004. C. Enz, F. Krummenacher, E. Vittoz, An analytical MOS transistor model valid in all regions of operation and

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

39

dedicated to low-voltage and low-current applications, J. Analog Int. Circ. Signal Proc. 8 (1995) 83-114. K. Allidina, M.N. El-Gamal, A 1 V CMOS LNA for low power ultra-wideband systems, in: IEEE ICECS 2008, Aug. 31-Sept. 3, 2008, pp. 165-168. A. Zolfaghari, B. Razavi, A low power 2.4-GHz transmitter/receiver CMOS IC, IEEE J. Solid-State Circuits 38 (2) (2003) 176-183. J.R. Clark, W.T. Hsu, M.A. Abdelmoneum, C.T.C. Nguyen, High-Q UHF micromechanical radial-contour mode disk resonators, IEEE J. Microelectromech. Syst. 14 (6) (2005) 1298-1310. A. Tekin, M.R. Yuce, W. Liu, Integrated VCO design for MICS transceivers, in: IEEE CICC 2006, 10-13 Sep., 2006, pp. 765-768. S. Ramasamy, B. Venkataramani, R. Niranjini, K. Suganya, 100 kHz-20 MHz programmable subthreshold Gm-C low-pass filter in 0.18 µm CMOS, VLSID 2009, pp. 105-110. B. Nauta, A CMOS transconductance-C filter for very high frequencies, IEEE J. Solid-State Circuits 27 (1992) 142-153. B. Guthrie, J. Hughes, T. Sayers, A. Spencer, A CMOS gyrator low-IF filter for a dual-mode Bluetooth/ZigBee transceiver, IEEE J. Solid-State Circuits 40 (9) (2005) 1872-1879. B.G. Perumana, C. Lee, S. Chakraborty, J. Laskar, Micro-power amplifiers for wireless PAN applications, in: Microwave Symposium Digest, IEEE MTT-S International, 12-17 June, 2005.


ANN Based Performance Analysis of UPFC in Power Flow Control S.K. Srivastava1 and S.N. Singh2 1. Department of Electrical Engineering, MMM Engineering College, Gorakhpur 273010, India 2. Department of Electrical Engineering, Indian Institute of Technology, Kanpur 208016, India

Received: January 15, 2010 / Accepted: February 09, 2010 / Published: August 31, 2010. Abstract: The Unified Power Flow Controller (UPFC) is one of the most versatile Flexible AC Transmission Systems (FACTS) devices that has unique capability of independently controlling the real and reactive power flows, in addition to regulate the system bus voltage. This paper presents performance analysis of Unified Power Flow Controller based on two axis theory. Based on this analysis, a new Artificial Neural Network (ANN) based controller has been proposed to improve the system performance. The controller rules are structured depending upon the relationship between series inserted voltage and the desired changes in real/reactive power flow in the power system. The effects of different controllers along with parameters of series transformer and transmission line have been investigated through developed control block model in SIMULINK tool box of MATLAB. The effectiveness of the proposed scheme is demonstrated by case studies. Key words: FACTS, UPFC, ANN, neural controller, SIMULINK.

1. Introduction Due to financial and environmental restrictions, utilities are trying to utilize the existing resources of the present day power system by maintaining the reliability and security levels. Flexible AC Transmission Systems (FACTS) [1, 2] have been introduced as means of high power semiconductor devices to rapidly control the power flow along a transmission corridor to reduce the power flows in heavily loaded lines resulting in increased system loadability, low system loss, improved stability of the network, reduced cost of production. The Unified Power Flow Controller (UPFC) is the most versatile and advance FACTS controller which can provide simultaneous real time control of all basic power system parameters (bus voltage, transmission line impedance, and transmission line angle), or any combination of these [2, 3]. Corresponding author: S.K Srivastava, associate professor, research fields: power system restructuring, FACTS, power system optimization and control, ANN & Fuzzy-Neural applications in power system. E-mail:[email protected].

The real and reactive power flows in transmission lines can be regulated by changing the magnitude and phase angle of series injected voltage produced by series inverter of UPFC. The shunt connected inverter provides the real power drawn by series branch and the losses. In addition, it can independently provide reactive power compensation to the system by controlling the reactive injection current. For simplification of control analysis and to improve the dynamic performance of UPFC, several control strategies including d-q axis control have been reported in the literature. Some of those have described the dynamic modeling of UPFC with conventional Proportional Integral (PI) and Proportional IntegralDifferential (PID) based control techniques [3, 4], whereas in Refs. [5, 6], fuzzy-rules based controllers of UPFC have been suggested to regulate the power system parameters and improve the dynamic performances. In Ref. [7], authors have proposed controllers for damping control and transient stability improvement through fuzzy logic based genetic algorithm,

ANN Based Performance Analysis of UPFC in Power Flow Control

P1, Q1

Rs Sending End

ise

Ls

Vse ∠θse

Vr ∠θr

V2 ∠θ2

Rse

Rr /2

Lse

Lr/2 Receiving End

P2, Q2

is ish

Lsh

41

Pref, Qref

Rsh Vsh∠θsh Fig. 1 Equivalent circuit model of UPFC.

which has resulted in parameters optimization with greater power flow control. In this paper, d-q axis control technique is extended for the control of UPFC. An attempt has made here to develop Artificial Neural Network (ANN) controller to replace the conventional one as proposed in Ref. [8]. A comprehensive model of UPFC based on d-q axis decoupled structure in SIMULINK block sets has been developed for neural network controller for both series and shunt controllers. The simulation results with case studies through developed neural network controller have been presented and compared those results with the conventional PI controller [9].

2. Dynamic Representation of UPFC A unified power flow controller consists of two back-to-back power electronics converters, which can control the real and reactive power flows in a transmission line. This control strategy is based on d-q axis theory. The dynamic model of UPFC is derived in the d-q (synchronously rotating at the system angular frequency ω) frame of Refs. [3, 10], and a ANN control strategy employed for active and reactive power control. The equivalent circuit model of a UPFC is shown in Fig. 1. The series and shunt converters are represented by controllable voltage sources Vse and Vsh, respectively. Rs and Ls are source resistance and

inductance, respectively; Rsh and Lsh are resistance and leakage inductance of the shunt transformer, respectively. Rse and Lse represent the resistance and leakage inductance of the series transformer, respectively. Rr and Lr are transmission line resistance and inductance respectively. ω is the angular frequency of the voltages and currents. Performing standard d-q transformation [8, 10] of the current through the shunt transformer and series transformer, various equations can be derived. 2.1 Shunt Converter Based Analysis From Fig.1, a mathematical model of the shunt converter of UPFC in the transmission system can be derived and is given as (V − Vshd ) dishd R = −ishd sh ω s + ω s ishq + 1d (1) dt Lsh Lsh (V1q − Vshq ) dishq R = −ishq sh ω s − ω s ishq + (2) dt L L sh

sh

Where the quantities with subscript d and q are d-axis and q-axis quantities, respectively. 2.2 Series Converter Based Analysis In order to control the real and reactive power flow at node 2, it is required to inject a series voltage of appropriate magnitude and phase angle. From Fig. 1, a mathematical model of the transmission system including series part of UPFC is derived as


42

(V + Vsed − Vrd ) dised R = −ised r + ω s iseq + 1d dt Lr Lr (V1q + Vseq − Vrq ) diseq R = −iseq r − ω s ised + dt Lr Lr

(3) (4)

2.3 DC Link Voltage Controller Analysis Net power input (real power input to shunt branch minus real power flow into series branch) to UPFC should instantaneously meet the charging rate of capacitor (C) and losses in the UPFC system. The mathematical equations can be derived and written as follows: ⎡ dV ⎤ Psh − Pse = Vdc ⎢C dc + g cVdc ⎥ (5) ⎣ dt ⎦ ⎡ dV ⎤ Vdc ⎢C dc + gcVdc ⎥ = Vshdishd + Vshqishq −Vsedised −Vseqiseq (6) ⎣ dt ⎦

or, dV dc gω 1 = c s V dc + V shd i shd + V shq i shq − V sed i sed − V seq i seq dt bc CV dc

[

] (7)

The dynamic behavior of the DC capacitor voltage can be realized by Eq. (7), where bc is the susceptance and gc is the conductance of the capacitor, respectively. DC voltage level is controlled by the shunt converter, which adjusts the amount of real power flow from the ac system into the common dc link.

3. ANN Based Control Algorithm In order to realize a real time control of the power flow in both steady and transient states, an ANN based control algorithm is proposed in this paper to control the power flow in transmission line. Instead of using the iteration method, a trained neural network can be used as an estimator to predict the internal control variables in the UPFC system. They mutually correspond to a given set of real and reactive power flows (P-Q) which is practically the real and reactive power flows controlled by UPFC. The main advantage of using ANN based estimators in the control scheme is that the time during the recall process of a trained estimator is almost negligible in comparison with the time taken to perform an iteration program to solve a set of non-linear equations [11] .

Neural networks are categorized by their architecture (layers), topology (connectivity pattern feed forward or recurrent, etc.) and learning regime. Most of the applications in power systems to date have used multi-layered feed forward networks of the type illustrated in Fig. 2 and use error back propagation learning [12, 13]. This technique adjusts the weights of the signals passing between neurons in order to produce an output which minimizes the total mean squared error of the network outputs when compared with a desired output response pattern. The trained ANN based predictor is able to correctly predict a set of proper control variables for the controllers to meet a certain goal [11, 14].

4. ANN Based Control Structure In this paper, feed forward control technique has been considered for controlling the parameters of UPFC. A feed forward neural network is a biologically inspired classification algorithm, consisting of a large number of simple neuron-like processing units, organized in layers. The back-propagation learning algorithm as emerged with feed forward multi-layer preceptron is simple to implement and computationally efficient in that its complexity is linear in the synaptic weights of the network. However, a major limitation of the algorithm is that it does not always converge and can be slow, particularly to deal with a difficult learning task that requires the use of a large network. The structure of most commonly used feed forward, multilayer back propagation-type network as shown in Fig. 3, consists of variable input signals. Each input signal flows through a gain or weight, called a synaptic weight or connection strength, whose function is analogous to that of the synaptic junction in a biological

Fig. 2 Schematic representation of a neuron.


43

Fig. 3 Structure of feed forward neural network. 1

TDL

weight

In1

Delays 1

IW{1,1}

2

TDL

weight

In2

Delays 2

IW{1,1}1

3

TDL

weight

In3

Delays 3

IW{1,1}2

1 purelin

Out1

bias b{1}

netsum

Fig. 4 Structure of neural controller.

neuron. The weights can be positive or negative, corresponding to “acceleration” or “inhibition”, respectively, of the flow of electrical signals in a biological cell. The summing node accumulates all the input weighted signals, adds the bias signal, and then passes to the output through the activation function, which is usually nonlinear in nature, as shown in the Fig. 3. Mathematically, the output expression can be given as ⎡ y = F(S) = F ⎢ ⎢⎣

N

∑

k =1

⎤ X kW k + b ⎥ ⎥⎦

such as

f i ( xi ) =

1 1 + e − bxi

or, f i (xi ) =

xi 1 + xi

(9)

(10)

The neural controller corresponding three input and weights with proper delay has been developed in SIMULINK as shown in Fig. 4.

5. SIMULINK Based Modeling of UPFC (8)

Denoting the weighted sum of input signals by Xi ,where Xi = W1X1 + W2 X2 + W3X3, then function fi(xi) may take typically one of many sigmoidal forms

Models for transmission line, shunt control, series control and DC voltage control are developed separately and then clubbed together to get the complete model of UPFC. The block diagram representation of SIMU-

44


LINK model of UPFC is shown in Fig. 5. The ANN based block model of series converter and shunt converter have been developed as shown in Figs. 6 and 7.

6. Simulation Results The performance of the developed block models in SIMULINK is evaluated with ANN controllers and compared with conventional PI controller. A 220 kV, 100 MVA, 150 km long line is considered for the simulation study. The active and reactive powers at the UPFC terminal towards the receiving end (P2, Q2) as shown in Fig. 1 are controlled to follow the commands ref

ref (Pref and Qref). The reference currents ised and i seq

are computed from equations as reported in Ref. [5]. The system parameters and controller parameters are given in Appendix.

settings in the active and reactive power flows are changed to 0.53 p.u and 0.52 p.u at 0.05 sec, which can be seen from Fig. 8. From the Fig. 8 , it is observed that both PI and ANN controller get settled to the steady state values but error in ANN controller is 0.1 pu. But the ANN controller gives smooth change to the steady state value compared to the PI controller. The ANN controller error can offset by adding the error signal to the output. Fig. 9 shows the reactive power flows with PI and ANN controllers. It can be seen that the ANN controller provides better settling time compared to the Shunt C o n v e rte r C o n tro l B lo c k

S o u rc e end m odel

D C L in k M odel

S e rie s C o n v e rte r C o n tro l B lo c k

The proposed control strategy has been tested with step change in active and reactive power flows. The simulation is started with no power flow (Vs=1∠0 and Vr=1∠0) and initial parameters setting of UPFC kept zero. After achieving the steady state, the reference

T ra n s m is s io n L in e M o d e l

Fig. 5 Block diagram representation of UPFC model.

Fig. 6 Block diagram representation of ANN based series converter.


45

Fig. 7 Block diagram representation of ANN based shunt converter. 0.8

0.6 PI ANN

0.6

0.5

0.4 Reactive Power(p.u)

Active power (p.u)

0.4

0.2

0

0.2

-0.2

0.1

-0.4

0

-0.6

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

PI ANN

0.3

-0.1

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

Fig. 8 Active power flow in the line.

Fig. 9 Reactive power flow in line.

PI controller. The voltage magnitude at bus-1, DC link voltage, shunt voltage magnitude, series injected voltage magnitude, voltage and angles of series injected voltage under ANN-PI control environment are shown in Figs. 10 to 15. From Fig. 12, it is found that ANN controller gives less capacitor voltage and therefore, the voltage rating of capacitor can be reduced and thus cost. The injected series voltage magnitude and phase angle is also less

with the ANN controller which shows that the rating of series transformer can be reduced. It gives a great saving in the cost of the total system.

7. Conclusions In this paper, the performance analysis of Unified Power Flow Controller (UPFC) based on two-axis theory has been presented using SIMULINK block set of MATLAB. The dynamic model of UPFC is derived


46

1.08

0.15 PI ANN

0.1

1.06 1.04

0.05

1.02

Vse(p.u)

Pse(p.u)

1

0

-0.05

0.98 PI ANN

0.96 0.94

-0.1 0.92

-0.15

0.9 0.88

-0.2

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0

0.05

0.1

0.15

0.2

0.5

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

Fig. 13 Series injected voltage (Vse).

Fig. 10 Series Converter power (Pse).

100 1.5 PI A NN

1.4

PI ANN

80 60

1.3

40 Angle(degree)

Vsh(p.u)

1.2 1.1 1

20 0 -20 -40

0.9

-60 0.8

-80 0.7

0

0.05

0.1

0.15

0.2

0.25 0.3 Tim e(s ec )

0.35

0.4

0.45

-100

0.5

Fig. 11 Shunt voltage magnitude (Vsh).

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

Fig. 14 Angles of series injected voltage.

1.25

1.4

PI ANN

PI ANN

1.3 1.2

1.2 Vdc(p.u)

Voltage V1(p.u)

1.1

1.15

1 0.9 0.8 0.7 0.6

1.1

0.5

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

0

0.05

0.1

0.15

0.2

0.25 0.3 Time(sec)

0.35

0.4

0.45

0.5

Fig. 12 DC link capacitor voltage (Vdc).

Fig. 15 Voltage magnitude (V1).

in the synchronously rotating d-q frame of reference and followed by ANN control strategy for real and reactive power control. The proposed ANN controller algorithms are structured depending upon the relation-

ship between series inserted voltage and the desired changes in real/reactive power flow in the powersystem. Simulation study reveals the following: (1) The ANN controller provides the quick response


with minimum overshoot compared to the PI controller. (2) The requirement of voltage rating of capacitor is reduced with ANN controller. (3) Series injected voltage magnitude and phase angle is less with ANN controller which shows that the rating of series transformer can be reduced. From the observation made, it can be concluded that ANN controller offers better regulating properties compared to the conventional PI controller. With the proposed ANN controller, the cost of UPFC can be reduced by having lower voltage rating of capacitor and reduced transformer rating of series injected voltage through the converter.

[6]

[7]

[8]

[9]

[10]

References [1]

[2]

[3]

[4]

[5]

L. Gyugui, Unified power flow concept for flexible AC transmission system, in: IEE Proceedings, part C, No.4, July, 1992, pp. 323-332. L. Gugui, C.D. Schauder, S.L. Williams, T.R. Rietman, D.R. Torgerson, A. Edris, The unified power flow controller, IEEE Transactions on Power Delivery 10 (2) (1995) 1085-1097. K.R. Padiar, A.M. Kulkarni, Control design and simulation of unified power flow controller, IEEE Transactions on Power Delivery 13 (4) (1997) 1346-1354. S.D. Round, Q. Yu, L.E. Norum, T.M. Underland, Dynamic control of a unified power flow controller, in: 27th Annual IEEE Power Electronics Specialists Conference, PESC’96, 1996, pp. 508-514. M.A.E. Farrag, G.A. Putrus, L. Ran, Design of fuzzy based-rules control system for the unified power flow

[11]

[12]

[13] [14]

47

controller, in: 28th Annual Conference of Industrial Electronics Society, IECON02, 5-8 Nov., 2002, Vol. 3, pp. 2102-2107. S. Limyingcharoen, U.D. Annakkage, N.C. Pahalawaththa, Fuzzy logic based unified power flow controllers for the transient stability improvement, IEE Proc-Gener. Transm. Distrib. 145 (3) (1998) 225-232. T.K. Mok, Y.X. Ni, F.F. Wu, Design of fuzzy damping controller of UPFC through genetic algorithm, in: IEEE PES Summer Meeting, USA, 2000, Vol. 3, pp. 1889-1894. O.P. Dwivedi, J.G. Singh, S.N. Singh, Simulation and analysis of unified power flow controller using simulink, in: National Power System Conference (NPSC-04), Dec., Vol. 2, 2004, pp. 1048-1054. R. Kumar, Comparative performance analysis of PI and neural network based controller for UPFC, M.Tech Thesis, 2007. S.K. Srivastava, K.G. Upadhyay, S. Khalid, S.N. Singh, Simulink based d-q axis model of UPFC, in: National Conference on Control, Communication and Information Systems, Goa, India, 23-24 Jan., 2004. K. Krishnaveni, G.T.R. Das, ANN based control patterns estimator for UPFC used in power flow problem, Journal of Theoretical and Applied Information Technology 3 (2007) 45-49. IEEE Tutorial on application of Artificial Neural Network to Power Systems (1996), IEEE catalog number – 96 (112) 74-88. L.Z. Yao, Neural Networks for Pattern Recognition, John Wiley & Sons Inc., 1993. K.L. Lo, T.T. Ma, J. Trecat, M. Crappe, A novel power flow control concept using ANN based multiple UPFC scheme, in: Proc. of Energy Management and Power Delivery, EMPD98, 3-5 Mar., 1998, Vol. 2, pp. 570-575.

Appendix: (1) System data The test system for simulation study of UPFC as shown in Fig.1 is single line circuit diagram model. The system data are as follows: ω0 = 2πf0, f0 = 50 Hz, Rs = 0.01 pu, Rr= 0.024 pu, Rsh = 0.04 pu, ω0Ls (=Xs) = 0.15 pu, ω0Lr (=Xr) = 0.12 pu, ω0Lsh (=Xsh) = 0.1 pu, Rse=0.01 pu, Xse=0.025, gsesh= 0.0067 pu, bsesh = 1.5708 pu, C = 5000 μF. (2) Controller data The parameters of the PI controllers are determined by a thorough and repeated study of the system response under various operating condition. The PI gains, which give the best responses under the tested conditions, as listed below: PI Controller

k psh1 = 0.5, k ish1 = 1000, k psh2 = 1.75, k ish2 = 2000 k psed = 0.5, k ised = 75, k pseq = 0.7, k iseq = 225


Spectral Calculation of the Output Voltage of an Inverter with Bipolar Pulse Width Modulation I. Mrčela, V. Šunde and Z. Benčić Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb HR-10000, Croatia Received: March 2, 2010 / Accepted: April 14, 2010 / Published: August 31, 2010. Abstract: Converters with pulse width modulation are used for connections between the direct current (DC) and alternating current (AC) networks, e.g., in uninterrupted power supply systems, AC electromotor drives, for powering induction furnaces, in audio technique. Spectrum of signals sampled by pulse amplitude modulation and output voltage spectrum of the converter with pulse width modulation have similar properties. Spectrum of signals sampled by pulse amplitude modulation contains a harmonic of frequency equal to the frequency of the modulating signal and the harmonics of frequencies equal to the sum of frequency of the modulating signal and multiples of the sampling frequency. The output voltage spectrum of the converter with bipolar pulse width modulation contains harmonic of frequency equal to the frequency of the modulating signal and harmonics of frequencies equal to sum of the frequency of the modulating signal and multiples of the frequency of the carrier signal. It also contains harmonics of frequencies equal to the sum of the multiples of the frequency of the modulating signal and the multiples of the carrier signal. The comparison analysis was carried out for the harmonics of the output voltage of the converter with bipolar pulse width modulation in time domain. The dependency of the amplitudes and frequency spectrum on the wave forms of the carrier signal and modulating signal was shown. Similarity of the output voltage spectrum of the converter and signal spectrum sampled by the pulse width modulation was also shown. Key words: Output voltage converter with bipolar pulse width modulation, spectral analysis, Fourier series, carrier signal, reference signal.

1. Introduction In Ref. [1] voltage spectrum calculation was carried out for the pulse width modulation converter (PWM-converter) using the Fourier transformation for a two variable function (double variable controlled waveform Fourier series). The calculation was performed by splitting of the voltage waveform into a Fourier series and by calculating the series coefficients, assuming a periodical voltage. The periodical voltage is essential for the foregoing applications. A periodical signal with the period T0 can be represented by Fourier series in trigonometric or in complex form. Trigonometric form of Fourier series is:

g (t ) = a 0 +

⎡

⎛ 2 ⋅ k ⋅π ⋅t ⎞ ⎟⎟ T0 ⎝ ⎠

∑ ⎢ a k ⋅ cos ⎜⎜

k =1 ⎣

⎛ 2 ⋅ k ⋅ π ⋅ t ⎞⎤ ⎟⎟ ⎥ + b k ⋅ sin ⎜⎜ T0 ⎝ ⎠⎦

Complex form of Fourier series is: ∞

g (t ) =

∑ cm ⋅ e

j⋅

2 ⋅k ⋅π ⋅t T0

(1)

(2)

m = −∞

If it can be assumed that the function g(t) is square integrable over the interval ⎡⎢− 0 , 0 ⎤⎥, coefficients a0, ⎣ 2 2⎦ ak, bk and ck are: T T

2 ⋅ T0

(3)

⎛ 2 ⋅ k ⋅π ⋅t ⎞ ∫ g (t ) ⋅ cos ⎜⎜⎝ T0 ⎟⎟⎠ ⋅ d t

(4)

T0 2 −

1 ⋅ T0

T0 2

∫ g (t ) ⋅ dt

a0 =

ak = Corresponding author: I Mrčela, junior researcher, research field: power electronics. E-mail: [email protected].

∞

T0 2

−

T0 2

Spectral Calculation of the Output Voltage of an Inverter with Bipolar Pulse Width Modulation

49

Fig. 1 Reference signal (modulating), carrier signal (transmitting signal) and modulated signal (sampled) in (a) pulse-amplitude modulation and (b) pulse-width modulation.

2 bk = ⋅ T0

T0 2

∫

−

T0 2

⎛ 2 ⋅ k ⋅π ⋅ t ⎞ ⎟⎟ ⋅ dt g (t ) ⋅ sin ⎜⎜ T0 ⎝ ⎠

1 1 ck = ⋅ (ak − j ⋅ bk ) = ⋅ 2 T0

T0 2

∫ g (t ) ⋅ e

−

− j⋅

2⋅k ⋅π ⋅t T0

(5)

⋅ dt (6)

T0 2

As an example, a spectral calculation was carried out for the reference signal of the sinusoidal waveform sampled by pulse amplitude modulation. Sampling of a signal by a series of the Dirac delta-functions gives the signal value (amplitude) for the moments of sampling. Sampling of a signal with pulse amplitude modulation can be mathematically modelled as a product of the modulating signal and carrier signal. Sampling of a signal with pulse width modulation gives the median value signal for the sampling period by the ratio of durations of individual positive and negative pulses. Sampling of a signal with pulse width modulation can be mathematically modelled as a sign function of the difference between the modulating signal and the carrier signal. Fig. 1(a) shows a block diagram of the sampling model for the sinusoidal signal by pulse amplitude modulation, and Fig. 1(b) shows a block diagram of the

sampling model of the sinusoidal signal by pulse width modulation. Modulation of the reference signal is, according to Ref. [2], carried out because the form of the modulated signal is suitable for transmission through the individual medium. The Fourier series for the signal modulated by pulse amplitude modulation is calculated as a product of the modulating signal Fourier series and the carrier signal Fourier series. The first step in the modulated signal spectrum calculation is calculation of the coefficients separation for the carrier signal into a Fourier series. The Fourier series in the complex form of Dirac delta-function δS(t) is:

δ S (t ) =

∞

∑ ck ⋅ e

j⋅

2⋅k ⋅π ⋅t TS

k = −∞

1 ∞ j⋅ = ⋅ ∑e TS k = −∞

2⋅k ⋅π ⋅t TS

(7)

Fourier series in complex form of the reference signal fr(t) = sin(ωr·t) is:

f r (t ) =

(

1 ⋅ e j⋅ω r ⋅t − e − j⋅ω r ⋅t 2⋅ j

)

(8)

The coefficients of the complex form Fourier series Eq. (8) are 1 and –1. The complex form Fourier series of the sinusoidal waveform signal sampled by pulse amplitude modulation is: f M (t ) = f r (t ) ⋅ δ S (t )

50


(

)

1 1 ∞ j⋅ = ⋅ e j⋅ω r ⋅t − e − j⋅ω r ⋅t ⋅ ⋅ ∑ e 2⋅ j TS k = −∞

2⋅ k ⋅π ⋅t TS

⎡ j⋅⎛⎜ ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ j⋅⎛⎜ −ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ ⎤ ∞ ⎜ r 1 ⎢ ⎜ r TS ⎟⎠ TS ⎟⎠ ⎥ = ⋅ ∑ ⎢e ⎝ −e ⎝ ⎥ 2 ⋅ j ⋅ TS k = −∞ ⎢⎣ ⎥⎦

⎡ ∞ j⋅⎛⎜ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ ∞ j⋅⎛⎜ −ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ ⎤ ⎜ r ⎜ r 1 ⎢ TS ⎟⎠ TS ⎟⎠ ⎥ = ⋅ ⎢ ∑e ⎝ − ∑e ⎝ ⎥ 2 ⋅ j ⋅ TS k =−∞ k =−∞ ⎢⎣ ⎥⎦ ⎡ ∞ j⋅⎛⎜ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ ∞ − j⋅⎛⎜ω ⋅t + 2⋅k ⋅π ⋅t ⎞⎟ ⎤ ⎜ r ⎜ r 1 ⎢ TS ⎟⎠ TS ⎟⎠ ⎥ = ⋅ ⎢ ∑e ⎝ − ∑e ⎝ ⎥ 2 ⋅ j ⋅ TS k =−∞ k =−∞ ⎥⎦ ⎢⎣ ⎛ ⎛ 2⋅k⋅π ⋅t ⎞ 2⋅k⋅π ⋅t ⎞ ⎤ ⎡ ⎟⎟ ⎟ − j⋅⎜⎜ωr ⋅t + j⋅⎜⎜ωr ⋅t + ∞ 1 ⎢ ∞ T TS ⎟⎠ ⎥ S ⎝ ⎠ ⎝ = ⋅ ∑(− j) ⋅ e + ∑j⋅ e ⎥ 2⋅TS ⎢k =−∞ k =−∞ ⎢⎣ ⎥⎦ (9) For the purpose of the Eq. (9) in trigonometric form, the real and imaginary parts of the coefficients in Eq. (9) must be separated. Real parts in both sums are zero, and the imaginary parts are –1 and 1. According to Eq.

(6),

it

follows

that

ck =

out for the cosine waveform of the reference signal, since that form is the most often used for the applications listed in the Abstract. The results of the spectral calculations for the modulated signal PWM-inverter were verified against the results of the calculations in Ref. [1]. The bipolar pulse width modulation is a two-level pulse width modulation, since the output voltage of the converter takes positive and negative values of the input voltage. For the interval where the reference signal is higher than the carrier signal, the converter output voltage is positive, and in the interval where the reference signal is lower than the carrier signal, the converter output voltage is negative. Fig. 2(a) shows the comparison of the reference signal fr(t) and the carrier signal fN(t), and Fig. 2(b) shows the modulated signal frm(t), as in Ref. [3]. The voltage waveform of the converter is equal to the waveform of the modulated signal. Time tS is the time shift of the carrier signal, and time tr is the time shift of

1 ⋅ (ak − j ⋅ bk ) and 2

according to Eq. (1), Eq. (9) in the trigonometric form:

f M (t ) =

⎛ 1 ⎡∞ 2 ⋅ k ⋅π ⋅ t ⎞ ⎟⎟ ⋅ ⎢ ∑ sin ⎜⎜ ωr ⋅ t + TS 2 ⋅ TS ⎢⎣ k =1 ⎝ ⎠

∞ ⎛ 2 ⋅ k ⋅ π ⋅ t ⎞⎤ ⎟⎟⎥ − ∑ sin ⎜⎜ − ωr ⋅ t − TS ⎝ ⎠⎥⎦ k =1

=

⎛ 1 ∞ 2 ⋅ k ⋅π ⋅ t ⎞ ⎟⎟ ⋅ ∑ sin ⎜⎜ ωr ⋅ t + TS k =1 ⎝ TS ⎠

(10)

The spectrum of the pulse amplitude modulation sampled signal is represented as a periodical spectrum of the sampled signal, with the period equal to the sampling period.

2. Bipolar Pulse Width Modulation for the Triangular Carrier Signal Spectral calculation was carried out for the harmonics of the modulated signal for the reference signal of a cosine waveform for the bipolar method of pulse width modulation. The calculation was carried

Fig. 2 (a) Comparison of the carrier signal fN(t) and the reference signal fr(t), (b) the modulated signal frm(t).

51


the reference signal. TS is the period of the carrier signal, Tr is the period of the reference signal. It is assumed that the phase shifts of the reference signal and the carrier signal are equal and that the ratio of TS and Tr is an integer number. If the ratio of TS and Tr is not an integer number, the modulated signal is not periodical and the Fourier transformation must be used instead of the Fourier series. The amplitudes of the reference signal and the carrier signal are taken to have value of 1, for the sake of simpler calculations of the Fourier series coefficients. The reference signal amplitude can take any value in the interval of [0, AN ], where AN is the amplitude of the carrier signal. Waveform of the modulated signal, as shown in Fig. 3(b), is periodical with the period of the reference signal Tr. Signal frm(t) in Fig. 3(a) can be represented as the sum of l signals frm(t)(1), ..., frm(t)(l), shown in Figs. T

r 3(b)-3(d), where l = T . Fourier series of the signal in S Fig. 3(a) can be represented as the sum of l Fourier series of the signals frm(t)(1), ..., frm(t)(l) in Figs. 3(b)-3 (d).

The coefficient ck of the signal frm(t)(1) of the complex form Fourier series is calculated according to Eq. (6). The integration boundaries are the crossing points of the reference signal and the carrier signal, i.e., solutions for the system of two equations: equations for the carrier signal and the reference signal. Periods T0 from Eq. (6) do not have to be equal, since the waveform signal in Fig. 3(a) consists of two periods: TS and Tr. If the integration limits are the solutions for the carrier signal equation, the exponent periods from Eqs. (6) and (2) are TS. If the integration limits are the solutions of the reference signal equations, the exponent periods from Eqs. (6) and (2) are Tr. Period T0 from Eq. (6) is the period of the signal in the Fig. 3(b) Tr for any form of solution of the equation system.

⎛ TS ⋅( f r (t −t r ) −1)+ tS ⎜ 1 4 ck (1) = ⋅ ⎜ e − j⋅2π ⋅k ⋅ f S ⋅t ⋅ dt + ∫ ⎜ Tr T ⎜ − S +t0 2 ⎝

TS + tS 2 − j⋅2π ⋅k ⋅ f S ⋅t

∫e

+

⋅ dt −

T − S ⋅( f r (t − t r ) −1)+ tS 4

⎞ ⎟ ⎟ ⋅ e d t ∫ ⎟ TS ⎟ ⋅( f r (t − t r ) −1)+ tS 4 ⎠ π ⎛ − j⋅ ⋅k ⋅( f r (t − t r ) −1)− j⋅2⋅π ⋅k ⋅ f S ⋅t S j ⋅ TS = ⋅⎜e 2 2π ⋅ k ⋅ Tr ⎜ ⎝ −

TS ⋅( f r (t − t r ) −1)+ t S 4 − j⋅2π ⋅k ⋅ f S ⋅t

− e j⋅π ⋅k − j⋅2⋅π ⋅k ⋅ f S ⋅tS + e − j⋅π ⋅k − j⋅2⋅π ⋅k ⋅ f S ⋅t S π j⋅ ⋅k ⋅( f r (t − t r ) −1)− j⋅2⋅π ⋅k ⋅ f S ⋅t S −e 2 − π j⋅ ⋅k ⋅( f r (t − t r ) −1)− j⋅2⋅π ⋅k ⋅ f S ⋅t S −e 2 π − j⋅ ⋅k ⋅( f r (t − t r ) −1)− j⋅2⋅π ⋅k ⋅ f S ⋅tS ⎞ ⎟ +e 2 ⎟ ⎠ 2 ⋅ TS π ⎤ ⎡ = ⋅ e − j⋅k ⋅ΘS ⋅ sin ⎢ ⋅ k ⋅ ( f r (t − tr ) − 1)⎥ (11) π ⋅ k ⋅ Tr ⎣2 ⎦ The signal time shift frm(t)(i) and frm(t)(i+1) is TS. The coefficient ck(i) of the signal frm(t)(i) is:

⎛ TS ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS ⎜ 1 ⎜4 ck (i ) = ⋅ e − j⋅ 2π ⋅ k ⋅ f S ⋅t ⋅ dt + ∫ ⎜ Tr T ⎜ − S + t S + (i −1)⋅TS − 2 ⎝ TS + t S + (i −1)⋅TS 2 − j⋅ 2π ⋅ k ⋅ f S ⋅t

∫e

T − S ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4

j ⋅ TS = 2π ⋅ k ⋅ Tr

−

⋅ dt −

TS ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4 − j⋅ 2π ⋅ k ⋅ f S ⋅t

∫e

TS ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4

⎞ ⎟ ⋅ dt ⎟ ⎟ ⎟ ⎠

⎛ − j⋅ π ⋅ k ⋅( f r (t − t r ) −1)− j⋅ 2⋅π ⋅ k ⋅ f S ⋅t S + (i −1)⋅TS ⋅⎜e 2 − ⎜ ⎝

− e j⋅π ⋅ k − j⋅ 2 ⋅π ⋅ k ⋅ f S ⋅t S + (i −1)⋅TS + e − j⋅π ⋅ k − j⋅ 2 ⋅π ⋅ k ⋅ f S ⋅t S + (i −1)⋅TS π

j⋅ ⋅ k ⋅( f r (t − t r ) −1)− j⋅ 2 ⋅π ⋅ k ⋅ f S ⋅t S + ( −e 2

+ (i −1)⋅TS

π

j⋅ ⋅ k ⋅( f r (t − t r ) −1)− j⋅ 2⋅π ⋅ k ⋅ f S ⋅t S + (i −1)⋅TS −e 2 π

− j⋅ ⋅ k ⋅( f r (t − t r ) −1)− j⋅ 2 ⋅π ⋅ k ⋅ f S ⋅t S + (i −1)⋅TS +e 2

⎞ ⎟ ⎟ ⎠

52


frm(t)

TS

frm(t)

t

a)

t

b)

Tr frm(t)(1) TS

Tr frm(t)(2) TS

t

Tr

c)

frm(t)(l) TS

t

Tr

d)

Fig. 3 In (b), (c) and (d) are parts of a modulated reference signal shown in (a) .

⎞ ⎟ = 2 ⋅ TS ⋅ e − j⋅ k ⋅ Θ S ⋅ e j⋅ 2π ⋅(i −1)⋅ k ⋅t ⋅ sin ⎡ π ⋅ k ⋅ ( f (t − t ) − 1)⎤ r r ⎢⎣ 2 ⎥⎦ ⎟ π ⋅ k ⋅T r ⎠ 2 ⋅ TS ⎡π ⎤ = ⋅ e − j⋅ k ⋅Θ S ⋅ sin ⎢ ⋅ k ⋅ ( f r (t − t r ) − 1)⎥ = ck (1) (12) 2 π ⋅ k ⋅ Tr ⎣ ⎦

ck(i) in Eq. (12) are functions depending on the waveform of the reference signal fr(t). The coefficients ck of the modulated reference signal are the sum of l coefficients ck(i):

cos(ωr·(t – tr)), for comparison of the results with the results in u [1], the coefficients ck are: ∞ 2 ⎛π ⎞ ck = ⋅ e − j⋅k ⋅Θ ⋅ ∑ J n ⎜ ⋅ k ⎟ ⋅ π ⋅k ⎝2 ⎠ n = −∞ π ⎛ ⎞ sin ⎜ (n − k ) ⋅ + n ⋅ (2 ⋅ π ⋅ f r ⋅ t − Θr )⎟ 2 ⎝ ⎠ ∞ 1 ⎛π ⎞ = ⋅ e − j⋅k ⋅Θ S ⋅ ∑ J n ⎜ ⋅ k ⎟ ⋅ j⋅π ⋅ k ⎝2 ⎠ n = −∞ S

⎛ j⋅⎛⎜ (n − k )⋅ π + n⋅(2⋅π ⋅ f r ⋅t −Θr )⎞⎟ ⎜ ⎝ 2 ⎠− ⎜e ⎜ ⎝

l

ck = ∑ ck (i ) = l ⋅ ck (1) = i =1

l⋅

2 ⎡π ⎤ ⋅ e − j⋅k ⋅ΘS ⋅ sin ⎢ ⋅ k ⋅ ( f r (t − tr ) − 1)⎥ π ⋅k ⋅l ⎣2 ⎦ 2 ⎡π ⎤ = ⋅ e− j⋅k ⋅ΘS ⋅ sin ⎢ ⋅ k ⋅ ( f r (t − tr ) − 1)⎥ (13) π ⋅k ⎣2 ⎦

If it is assumed that the reference signal is fr(t – tr) =

π ⎛ ⎞ − j⋅⎜ (n − k )⋅ + n⋅(2⋅π ⋅ f r ⋅t − Θr )⎟ ⎞ 2 ⎝ ⎠⎟ −e

=

1 j⋅π ⋅ k

⋅ e − j⋅k ⋅ΘS ⋅

∞

⎟ ⎟ ⎠

⎛π ⎞ Jn ⎜ ⋅ k ⎟ ⎝2 ⎠ n = −∞

∑

53


⎛ j⋅(n − k )⋅ π ⎜e 2 ⋅ e j⋅(n⋅ω r ⋅t − n⋅Θr ) − ⎜ ⎝ π ⎞ − j⋅(n − k )⋅ 2 ⋅ e − j⋅(n⋅ω r ⋅t − n⋅Θr ) ⎟ −e ⎟ ⎠

− j ⋅ sin (− n ⋅ ωr ⋅ t + n ⋅ Θr )]}⋅ e j⋅(k ⋅ωs ⋅t − k ⋅ΘS )

The sine is an odd and the cosine is an even function. The following applies: ∞

(14)

Where Θr, a product of tr and ωr, is the phase shift of the reference signal, and ΘS, a product of tS and ωS, is the phase shift of the carrier signal. Jn are Bessel functions of the first kind of the n-th order, and fr is the frequency of the reference signal. Instead of calculating ck for all cases of the integer n and k = 0, a0 can be calculated according to Eq. (3), after representing the Fourier series in trigonometric form. By inserting Eq. (14) into Eq. (2), the modulated signal can be represented as the Fourier series in complex form: f rm (t ) = c0 +

k = −∞

∞

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) − cos(− n ⋅ ωr ⋅ t + n ⋅ Θr ) =

n = −∞ ∞

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) − cos[− (n ⋅ ωr ⋅ t − n ⋅ Θr )]

=

n = −∞ ∞

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) − cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) = 0

n = −∞

(17) ∞

∑ sin (n ⋅ ωr ⋅ t − n ⋅ Θr ) + sin (− n ⋅ ωr ⋅ t + n ⋅ Θr ) =

n = −∞ ∞

∑ sin (n ⋅ ωr ⋅ t − n ⋅ Θr ) + sin[− (n ⋅ ωr ⋅ t − n ⋅ Θr )]

=

n = −∞ ∞

∑ sin(n ⋅ ωr ⋅ t − n ⋅ Θr ) − sin(n ⋅ ωr ⋅ t − n ⋅ Θr ) = 0

n = −∞

1 1 ⎛π ⎞ ⋅ e − j⋅k ⋅Θ S ⋅ ∑ ∑ ⋅ J n ⎜ ⋅ k ⎟ ⋅ j⋅π k ⎝2 ⎠ k = −∞ n = −∞

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) + cos(− n ⋅ ωr ⋅ t + n ⋅ Θr )

) = ∑ cos(n ⋅ ω r ⋅ t − n ⋅ Θr ) + cos[− (n ⋅ ω r ⋅ t − n ⋅ Θr )]

⎛ j⋅(n − k )⋅ π ⎞ j⋅(k − n )⋅ ⎜e 2 ⋅ e j⋅(n⋅ω r ⋅t − n⋅Θr ) − e 2 ⋅ e − j⋅(n⋅ω r ⋅t − n ⋅Θr ) ⎟ ⋅e j⋅k ⋅ω s ⋅t ⎜ ⎟ ⎝ ⎠

n = −∞ ∞

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) + cos(n ⋅ ωr ⋅ t − n ⋅ Θr )

=

n = −∞

k = −∞ ∞ 1 1 ⎛π ⎞ ⎧ ⋅ e j⋅k ⋅Θ S ⋅ ∑ ∑ ⋅ J n ⎜ ⋅ k ⎟ ⋅ ⎨ j⋅π ⎝2 ⎠ k = −∞ n = −∞ k

⋅e

⎬⋅e ⎭

(15)

Eq. (15) can be written as: f rm (t ) = c0 +

∑ cos(n ⋅ ωr ⋅ t − n ⋅ Θr )

n = −∞

∞

⎧⎡ ⎛ π⎞ π ⎞⎤ j⋅(n⋅ω r ⋅t − n⋅Θr ) ⎛ − ⎨⎢cos⎜ (n − k ) ⋅ ⎟ + j ⋅ sin ⎜ (n − k ) ⋅ ⎟⎥ ⋅ e 2 2 ⎠⎦ ⎝ ⎠ ⎝ ⎩⎣ ⎡ ⎛ π⎞ π ⎞⎤ ⎛ − ⎢cos⎜ (n − k ) ⋅ ⎟ − j ⋅ sin ⎜ (n − k ) ⋅ ⎟⎥ ⋅ 2 2 ⎠⎦ ⎝ ⎝ ⎠ ⎣ j⋅(k ⋅ω s ⋅t − k ⋅Θ S )

∞

= 2⋅

k ≠0

− j⋅(n⋅ω r ⋅t − n⋅Θr ) ⎫

(18)

∞

n = −∞ ∞

k ≠0 π

= c0 +

(16)

∑ sin(n ⋅ ωr ⋅ t − n ⋅ Θr ) − sin(− n ⋅ ωr ⋅ t + n ⋅ Θr )

=

n = −∞ ∞

∑ sin (n ⋅ ωr ⋅ t − n ⋅ Θr ) − sin[− (n ⋅ ωr ⋅ t − n ⋅ Θr )]

n = −∞ ∞

=

∑ sin (n ⋅ ωr ⋅ t − n ⋅ Θr ) + sin(n ⋅ ωr ⋅ t − n ⋅ Θr )

n = −∞

∞

= 2⋅

∑ sin (n ⋅ ωr ⋅ t − n ⋅ Θr )

n = −∞

1 k = −∞ ∞ 1 ⋅ ∑ ∑ ⋅ j ⋅ π k = −∞ n = −∞ k k ≠0

π⎞ ⎛π ⎞ ⎧ ⎛ J n ⎜ ⋅ k ⎟ ⋅ ⎨cos⎜ (n − k ) ⋅ ⎟ ⋅ 2⎠ ⎝2 ⎠ ⎩ ⎝ [cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) + j ⋅ sin(n ⋅ ωr ⋅ t − n ⋅ Θr )

− cos (− n ⋅ ω r ⋅ t + n ⋅ Θr ) + j ⋅ sin (− n ⋅ ω r ⋅ t + n ⋅ Θr )] +

π⎞ ⎛ + j ⋅ sin⎜ (n − k ) ⋅ ⎟ ⋅ [cos(n ⋅ ωr ⋅ t − n ⋅ Θr ) + 2⎠ ⎝ + j ⋅ sin (n ⋅ ωr ⋅ t − n ⋅ Θr ) + cos(− n ⋅ ωr ⋅ t + n ⋅ Θr ) −

(19)

(20)

It can be written:

f rm (t ) = c0 +

∞ ∞ 1 2 ⎛π ⎞ ⋅ ∑ ∑ ⋅ Jn⎜ ⋅ k ⎟ ⋅ j⋅ j ⋅ π k = −∞ n = −∞ k ⎝2 ⎠ k ≠0

π⎞ ⎛ sin⎜ (n − k ) ⋅ ⎟ ⋅ e j⋅(n⋅ωr ⋅t − n⋅Θr ) ⋅ e j⋅(k ⋅ωs ⋅t − k ⋅ΘS ) 2⎠ ⎝ ∞ ∞ 2 ⎛π ⎞ = c0 + ∑ ∑ ⋅ Jn ⎜ ⋅ k ⎟ ⋅ ⎝2 ⎠ k = −∞ n = −∞ k ⋅ π k ≠0

π⎞ ⎛ sin ⎜ (n − k ) ⋅ ⎟ ⋅ e j⋅[n⋅(ωr ⋅t −Θr )+ k ⋅(ωs ⋅t − ΘS )] (21) 2⎠ ⎝

54


Resolution of real and imaginary parts of the coefficients in Eq. (21) due to the Eq. (21) in the trigonometric form is carried out analogously to those in the section Introduction. The Eq. (21) in the trigonometric form is: ∞

∞

π⎞ 4 ⎛ ⎛π ⎞ ⋅ J n ⎜ ⋅ k ⎟ ⋅ sin⎜ (n − k ) ⋅ ⎟ 2⎠ ⎝ ⎝2 ⎠ k =1 n = −∞ k ⋅ π

f rm (t ) = a0 + ∑

∑

cos[n ⋅ (ωr ⋅ t − Θr ) + k ⋅ (ωs ⋅ t − ΘS )]

(22)

The coefficient a0 is calculated according to Eq. (3), analogously to calculation of the coefficients ck: ⎛ TS ⋅( f r (t −t r ) −1)+ tS + (i −1)⋅TS ⎜ 1 ⎜4 a0 (i ) = ⋅ dt + ∫1⋅ Tr ⎜ TS ⎜ − + t S + (i −1)⋅TS 2 ⎝ TS + t S + (i −1)⋅TS 2

+

t−

∫1⋅

dt −

∫1

⎞ ⎟ ⋅ dt ⎟ ⎟ ⎟ ⎠

T − S ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4 TS − ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4 TS ⋅( f r (t − t r ) −1)+ t S + (i −1)⋅TS 4

=

TS ⋅ cos(ωr ⋅ t − Θr ) Tr

(23) (24)

i =1

By insertion of Eq. (24) into Eq. (22), the modulated signal can be represented as: ∞

∞

neglected. The harmonics of the modulated signal are shown in the Fig. 4, for the frequency of the carrier signal fS = 2000 Hz and the frequency of the reference signal of fr = 50 Hz. The frequency of the carrier signal is dominant in comparison to the frequency of the reference signal and the harmonics k ≠ 0 and n ≠ 0 from Eq. (13) are repeated every k·fS, as in the Fig. 4. Harmonics for k ≠ 0 and n ≠ 0 are called carrier and sideband harmonics in Ref. [1]. Carrier and sideband harmonics are a consequence of two frequencies contained in the waveform of the voltage of the PWM-converter frequency of the modulating signal and carrier signal.

3. Bipolar Pulse Width Modulation in the Case of Sawtooth Carrier Signal

l

a0 = ∑ a0 (i ) = cos(ωr ⋅ t − Θr )

second order, for the same argument. In Ref. [1] the amounts for the Bessel function for n > 9 are

f rm (t ) = cos (ωr ⋅ t − Θr )

The calculation of the Fourier series for modulated signal by bipolar pulse width modulation in the case of a sawtooth carrier signal waveform was carried out in the same manner as in the section for the bipolar modulation of the pulse width in the case of triangular carrier signal waveform. Coefficients ck(i) for the modulated signal Fourier series frm(t) are:

π⎞ 4 ⎛π ⎞ ⎛ ⋅ J n ⎜ ⋅ k ⎟ ⋅ sin ⎜ (n − k ) ⋅ ⎟ ⋅ 2⎠ ⎝2 ⎠ ⎝ k =1 n = −∞ k ⋅ π cos[n ⋅ (ωr ⋅ t − Θr ) + k ⋅ (ωS ⋅ t − ΘS )] (25)

+∑

∑

The Bessel function of the first order for the argument

π

2

is the form: n + 2 ⋅l

⎛π ⎞ ⎜ ⎟ 2 ⎛π ⎞ J n ⎜ ⋅ k ⎟ = ∑ (− 1)l ⋅ n +⎝2⋅l ⎠ 2 l!⋅2 ⋅ Γ(n + l + 1) ⎝ ⎠ l =0 ∞

(26)

By growth of order of the Bessel function n, its value decreases exponentially. The Bessel function of the ninth order is by five orders of magnitude smaller in comparison to the Bessel function of the first and the

Fig. 4 Amplitude - frequency properties of the PWM-converter output voltage in the case of the triangular carrier signal.

55


⋅ e j⋅[k ⋅(ω s ⋅t − Θ S )− n ⋅(ω r ⋅t − Θ r )] − ∞ j ∑ π ⋅ k ⋅ e j⋅k ⋅π ⋅ e j⋅k ⋅(ωs ⋅t − Θ S ) k = −∞

⎛ TS ⋅ f r (t − t r ) + (i −1)⋅TS + t S ⎜2 1 ck (i ) = ⋅ ⎜ e − j⋅ 2π ⋅ k ⋅ f S ⋅t ⋅ dt − ∫ ⎜ Tr ⎜ − TS + t S + (i −1)⋅TS 2 ⎝

⎞ ⎟ − ⋅ dt ⎟ ∫e ⎟ TS ⎟ ⋅ f r ( t − t r ) + (i −1)⋅TS + t S 2 ⎠ j ⋅ TS = ⋅ e − j⋅ k ⋅Θ S ⋅ e − j⋅π ⋅ k ⋅( f r (t − t r ) ) − e j⋅ k ⋅π π ⋅ k ⋅ Tr

k ≠0

= c0 +

TS + t S + (i −1)⋅TS 2 − j⋅ 2π ⋅ k ⋅ f S ⋅t

(

j⋅ n ⋅

)

j ⋅ TS = ⋅ e − j⋅ k ⋅ Θ S π ⋅ k ⋅ Tr

π⎞ ⎛ ⎛ ∞ − j⋅ ⎜ n ⋅ (ω r ⋅ t − Θ r )+ n ⋅ ⎟ ⎜ 2⎠ ⎝ − e j⋅ k ⋅π ⎜ ∑ J n (π ⋅ k ) ⋅ e ⎜ n = −∞ ⎝

⎞ ⎟ ⎟ (27) ⎟ ⎠

Coefficients ck are: l

ck = ∑ ck (i ) = l ⋅ ck (i ) = l ⋅ i =1

j ⋅ TS ⋅ e− j⋅k ⋅ΘS ⋅ π ⋅ k ⋅ Tr

π⎞ ⎛ ⎞ ⎛ ∞ − j⋅⎜ n⋅(ω r ⋅t − Θ r )+ n⋅ ⎟ ⎜ j⋅k ⋅π ⎟ 2⎠ ⎝ ( ) J π k e e ⋅ ⋅ − ⎟ ⎜ ∑ n ⎟ ⎜ n = −∞ ⎠ ⎝ j = ⋅ e − j⋅k ⋅ΘS ⋅ π ⋅k π⎞ ⎛ ⎛ ∞ ⎞ − j⋅⎜ n⋅(ω r ⋅t − Θ r )+ n⋅ ⎟ ⎜ j⋅k ⋅π ⎟ 2⎠ ⎝ −e ⎜ ∑ J n (π ⋅ k ) ⋅ e ⎟ (28) ⎜ n = −∞ ⎟ ⎝ ⎠

Instead of calculating the coefficient c0, the coefficient a0 can be calculated according to Eq. (3), after the Fourier series is represented in trigonometric form. By inserting Eq. (28) into Eq. (2), the modulated signal it can be represented as a Fourier series in complex form:

f rm (t ) = c0 +

⎡ ∞ j ∑ ⎢⎢ ∑ π ⋅ k ⋅ J n (π ⋅ k ) ⋅ k = −∞ ⎢ n = −∞ k ≠0 ⎣ ∞

π⎞

e

⎛ − j⋅⎜ n ⋅(ω r ⋅t − Θ r )+ n ⋅ ⎟ 2⎠ ⎝

= c0 +

∞

∞

⎤ − e j⋅ k ⋅π ⎥ ⋅ e j⋅ k ⋅(ω s ⋅t − Θ S ) ⎥ ⎥⎦ π

− j⋅ n ⋅ j ∑ ∑ π ⋅ k ⋅ J n (π ⋅ k ) ⋅ e 2 ⋅ k = −∞ n = −∞ k ≠0

π

∞

∞

j ⋅ J n (π ⋅ k ) ⋅ k = −∞ n = −∞ π ⋅ k

∑ ∑

k ≠0 n≠0

e 2 ⋅ e j⋅[k ⋅(ω s ⋅t − Θ S )− n ⋅(ω r ⋅t − Θ r )] + ∞ j ∑ π ⋅ k ⋅ e j⋅k ⋅(ωs ⋅t − Θ S ) ⋅ J 0 (π ⋅ k ) − e j⋅k ⋅π k = −∞

(

) (29)

k ≠0

Representation of the Fourier series in trigonometric form is done analogously to the case of the triangular carrier signal. Coefficient a0(i) is:

⎛ TS ⋅ f r (t − t r ) + (i −1)⋅TS + tS ⎜ 1 2 a0 (i ) = ⋅ ⎜ ⋅ dt − ∫1 Tr ⎜ TS ⎜ − + tS + (i −1)⋅TS 2 ⎝ TS ⎞ + t S + (i −1)⋅TS ⎟ 2 T ⋅ dt ⎟ = S ⋅ f r (t − t r ) (30) ∫1 ⎟ Tr TS ⎟ ⋅ f r (t − t r ) + (i −1)⋅TS + t S 2 ⎠ Coefficient a0 is: l

a0 = ∑ a0 (i ) = f r (t − tr ) = cos(ωr ⋅ t − Θr )

(31)

i =1

The modulated signal represented as a Fourier series in trigonometric form is: ∞ ∞ 2 f rm (t ) = a0 + ∑ ∑ ⋅ J n (π ⋅ k ) k = −∞ n = −∞ π ⋅ k k ≠0 n≠0

⎧ ⎛ π⎞ ⎨sin ⎜ n ⋅ ⎟ ⋅ cos[k ⋅ (ωs ⋅ t − ΘS ) − n ⋅ (ωr ⋅ t − Θ r )] − ⎩ ⎝ 2⎠ ⎫ ⎛ π⎞ − cos⎜ n ⋅ ⎟ ⋅ sin[k ⋅ (ωs ⋅ t − ΘS ) − n ⋅ (ωr ⋅ t − Θ r )]⎬ ⎝ 2⎠ ⎭ ∞ 2 +∑ ⋅ (cos(m ⋅ π ) − J 0 (π ⋅ k )) ⋅ sin (ωs ⋅ t − ΘS ) k =1 π ⋅ k ∞ ∞ 2 = cos(ωr ⋅ t − Θ r ) + ∑ ∑ ⋅ J n (π ⋅ k ) ⋅ π k = −∞ n = −∞ ⋅ k k ≠0 n≠0

⎧ ⎛ π⎞ ⎨sin ⎜ n ⋅ ⎟ ⋅ cos[k ⋅ (ωs ⋅ t − Θ s ) − n ⋅ (ω r ⋅ t − Θ r )] − ⎩ ⎝ 2⎠ ⎫ ⎛ π⎞ − cos⎜ n ⋅ ⎟ ⋅ sin[k ⋅ (ωs ⋅ t − Θs ) − n ⋅ (ω r ⋅ t − Θ r )]⎬ + ⎝ 2⎠ ⎭ ∞ 2 +∑ ⋅ [cos (m ⋅ π ) − J 0 (π ⋅ k )] ⋅ sin (ωs ⋅ t − Θ s ) (32) k =1 π ⋅ k

56


Spectrum of the modulated signal sampled by pulse width modulation in case of a sawtooth carrier signal differs from the spectrum of the modulated signal sampled by a triangular carrier signal in the contents of the accompanying harmonics. This is a consequence of changed integration boundaries while calculating Fourier series’ coefficients ck. As well as in the case of pulse width modulated with a triangular carrier signal, harmonics of the modulated signal have a frequency equal to the sum of the multiples of the frequency of the carrier signal and the multiples of the reference signal.

controlled waveform Fourier series. Changing the boundaries of integration while calculating the members of the Fourier series applies to the procedure for calculation of the frequency spectrum for any form of the carrier signal and the reference signal. It was demonstrated that the signal spectrum sampled by various forms of the carrier signals differ in the contents of the carrier and sideband harmonics. Appearance of the carrier and sideband harmonics in the output voltage spectrum of the converter is the result of two frequencies contained in the waveform of the converter voltage.

4. Conclusions

References

The analysis of the frequency harmonics of the converter output voltage with the bipolar pulse width modulation was carried out in time domain. The applied calculation method for the spectrum is performed without a need to know the double variable

[1]

[2] [3]

D.G. Holmes, T.A. Lipo, Pulse Width Modulation for Power Converters – Principles and Practice, IEEE Press, Wiley InterScience, 2003. Technical Lexicon, Miroslav Krleža Lexicografical Institute, 2007, (in Croatian). J.G. Kassakian, Principles of Power Electronics (in Croatian), Graphis, Zagreb, 2000.


The Energy Balance of the Electro-Hydraulic Linear Actuation System M. Herranen, K. Huhtala and M. Vilenius Dept. of Intelligent Hydraulics and Automation, Tampere University of Technology, Tampere 33101, Finland

Received: March 17, 2010 / Accepted: April 22, 2010 / Published: August 31, 2010. Abstract: The variable gas exchange valve actuation systems have been developed in order to improve the efficiency of the combustion process. The electro-hydraulic valve actuation (EHVA) systems have good power to weight ratio, high maximum force and good controllability. The disadvantages are limited frequency bandwidth and energy recovery. Each component of the EHVA system has certain energy consumption, which is characteristic to the component. In this study the power consumptions of the components are investigated by means of the simulation. The investigated components are a hydraulic pump, a hydraulic accumulator, a control valve, and hydraulic lines connecting the components. The pressure losses caused by the oil flow are most significant in the control valves, 50-60% of the total energy consumption. If the stored kinetic energy of the actuator and moving oil masses could be reused, the energy consumption could be up to 25% better. Key words: Energy, consumption, electro-hydraulic, variable valve train.

Nomenclature A aact B Ehyd open Ehyd total Enom open Fact P ppump p+ Q+ T topen tstroke vact

Actuator port of the hydraulic control valve Acceleration of the actuator Actuator port of the hydraulic control valve Hydraulic energy needed to open the gas exchange valve Hydraulic energy during the stroke produced by the pump Energy needed to accelerate the actuator Force affecting to the actuator Pressure port of the hydraulic control valve Pressure in pump pressure line Pressure in the upper cavity of the actuator Hydraulic flow in or out the upper cavity of the actuator Tank port of the hydraulic control valve Time of the open event Time of the stroke event Speed of the actuator

tems have fully flexible opening time and lift events. Thus the fuel efficiency of the combustion process can be controlled better under the varying load conditions. The EHVA system can be a useful device to reduce the emissions too. However, the total energy consumption shouldn’t be much higher than the conventional cam driven valve actuation system. In Fig. 1, the example of the friction distribution of the conventional engine is shown. In the EHVA system has no cam/follower contact or the camshaft journals, which reduces the friction significantly [1].

1. Introduction The electro-hydraulic valve actuation (EHVA) sysCorresponding author: M. Herranen, M.Sc., researcher, research fields: diesel engine hydraulics, hydraulic system simulations. E-mail: [email protected].

Fig. 1

Valve train friction loss percentage [1].

58

The Energy Balance of the Electro-Hydraulic Linear Actuation System

Until now, most of EHVA studies have concentrated to the properties or function of the actuation system. Accurate comparison of the energy balance or the energy consumption has been rarely published. The investigated EHVA system can be built without the return spring of the gas exchange valve, which could change the energy balance as well. However, the hydraulic control valve consumes the electric energy and this will add the EHVA total energy consumption. Also the pressure drop caused by the flow of the hydraulic fluid decreases the efficiency. The energy taken by the hydraulic pump is dependent on the flow rate and pressure. If the pressure or the flow rate could be taken down while no movements are needed, energy saving would be considerable [2]. In the EHVA system the potential energy stored in the pressurized fluid is converted to the kinetic energy of the gas exchange valve system. The problem in the EHVA system is how the kinetic energy could be recovered to the pressure energy.

2. EHVA System Preliminary energy consumption calculation of the different hydraulic systems has been made in earlier study [3]. In this paper, the previously chosen best hydraulic circuit with return spring is taken under investigation. Comparable hydraulic system without the spring had worse energy consumption, but it could have some other benefits, and is therefore chosen. 2.1 Hydraulic Circuits In Fig. 2 and Fig. 3 the simplified schematic pictures of the studied hydraulic systems are shown. In Fig. 2, the constant hydraulic pressure is affecting to the circular area in the bottom of the actuator. When the pressurized fluid is directed to the above of the actuator, the actuator opens the gas exchange valve and compresses the return spring. The circular area also pushes the pressurized fluid to the accumulator. When the upper chamber is connected to the tank, the spring and the pressurized circular area closes the gas exchange

Fig. 2 Hydraulic circuit of the return spring system.

Fig. 3 Hydraulic circuit of the system without return spring.

valve. The circular area helps to move only the actuator faster; it can not draw the gas exchange valve to close direction. In another hydraulic circuit (Fig. 3) the pressure of the circular area is controlled also, and no return spring is applied. The opening and closing of the gas exchange valve is done only by means of hydraulic pressure forces. The closing sequence is performed by actuator only via the fixed coupling between the actuator and gas exchange valves. 2.2 Hydraulic Components Each hydraulic component has a certain efficiency which may be originated from mechanical and flow friction, compressibility of the medium, and leakages. Some components need the electric power in order to operate. All these energy sinks degrade the total efficiency of the EHVA system.

The power consumptions of the hydraulic control valve and the hydraulic line between the valve and the actuator are investigated. The hydraulic pump is sized by the average flow demand, but it must be oversized slightly in order to compensate the leakages and to maintain the pressure. The additional flow is directed through the pressure relief valve to the tank, and is producing pure waste energy. The amount of extra flow is depended on the type of the pump and the hydraulic circuit. The variable displacement pump can adjust the changes in average flow rate. The comparison between two pump types is done. The momentary flow demand of the system is much higher than average flow. It’s not reasonable to size the fixed hydraulic pump according to max. flow, and the change is faster than variable pump can follow, so the extra flow peak is utilized either the hydraulic/gas accumulator or the compressed fluid volume.

3. Simulations

59

Actuator lift [m]


Fig. 4 Stroke event of the actuator.

the pump (Eq. (2)), and the relational energy consumption is calculated.

(1)

(2) 3.1 Hydraulic Circuit Comparison In the system with the return spring, the energy stored in the spring produces the return stroke, and

The simulations are made by the AMESim (LMS International) program which is complete 1D simulation suite to model and analyze multi-domain, intelligent systems, and predict their multi-disciplinary performance [4]. The flow and pressure are investigated before and after every component. Due to the differences between the hydraulic systems, the control parameters had to be tuned in order to achieve closely identical displacement curves. The stroke event is shown in Fig. 4. The effect of the cylinder pressure is modeled, too. The hydraulic energy taken during the stroke is calculated by integrating the multiply of the hydraulic pressure and flow rate to the actuator chamber. This is the amount of energy which is eventually needed to produce the stroke of the electro-hydraulic gas exchange valve system. The energy is calculated separately in open (Eq. (1)) and close direction. After that the energy loss/consumption over any component is compared to the total hydraulic energy produced by

therefore, any extra closing energy is futile. However, the actuator has to push out the hydraulic fluid from above of the actuator and this needs some energy. The returning flow adds the energy need in all cases. The energy of the opening and closing events include the energy needed to overtake energy losses in returning hydraulic lines and components. Without the return spring, the actuator has to draw whole gas exchange valve system to the upper position, and therefore the energy consumption is much higher. The needed energy levels are presented in Table 1. The difference in overall energy rate is explained by higher flow rate of the non-return spring system. 3.2 Hydraulic Components 3.2.1 Control Valve The hydraulic control valve has significant effect to the system efficiency. In Table 2 there is shown the relatively energy losses over the control valve, compared

60


Table 1 Energy rates of the systems. Energy Opening (hydraulic) Closing (hydraulic) Pressure relief valve Energy from the pump System efficiency

Return spring 179 J 28 J 2J 309 J 67 %

No spring 132 J 116 J 12 J 410 J 60 %

Table 2 Energy loss in the control valve. P-->A A-->T B-->T P-->B Total

Return spring (%) 35 23 58

No spring (%) 21 % 27 % 8% 10 % 66 %

Fig. 5 Energy consumption in hydraulic line (control valve ↔ actuator).

to the energy produced by the pump during the event. 3.2.2 Hydraulic Lines The length of the hydraulic lines between the components can produce some energy losses. The length of the line has straight effect to the acceleration pressure force needed in the beginning of the stroke. The l/d ratio has effect to the viscous friction of the fluid flow. In the simulations, the basic pipeline length between the control valve and the actuator is 0.2 m and diameter 10 mm. Then the length of the line is altered. In Fig. 5 the energy consumption during the stroke in that specific hydraulic line is presented. The energy consumption is increased linearly when the l/d-ratio increases. As a percentage the single line energy consumption is under 1% from the total energy consumed. The energy consumption is higher in lines between the control valve and pump/tank, because

higher l/d ratio (length of the pipe 2 m, diameter 19 mm). Total energy consumption of the pipe lines can be assumed to be around 7%-8% in both investigated hydraulic circuits. The flow rate without the return spring is losing less energy in single line, but total energy consumption is nearly the same because the flow goes in parallel to two lines during the stroke. The non-return spring system has higher total flow rate from the pump, and energy consumption between the pump and the control valve is higher. The return spring system has, however, high flow rate peak to the tank line while the non-return spring system the flow is more evenly distributed during the stroke. This also balances the difference between the systems. 3.2.3 Hydraulic Pump The hydraulic pump should produce only required amount of pressurized fluid. The fixed displacement pump takes all the time nearly fixed amount of energy, when running. If the gas exchange valve lift is decreased, the rise of the waste energy directly corresponds the decreased actuator lift (Table 3). The relatively efficiency decreases, but so does the difference between the systems. This is partially due to control valve losses, which are now 28% (return spring) and 31% (non-return spring cases) from the total energy losses. With the variable displacement pump, the EHVA system efficiencies would be 68% (return spring) and 66% (non-return spring) though the actuator lift would be reduced. The controller of the variable displacement pump takes some energy when the displacement is altered. This, however, is not taken into account in this study. Also must keep in mind that the efficiency of the pump will decrease, when partial displacement is used. The reduction of the pump Table 3 Energy rates of the systems, 50% actuator lift, fixed pump. Energy Opening (hydraulic) Closing (hydraulic) Pressure relief valve Energy from the pump System Efficiency

Return spring 98 J 15 J 160 J 320 J 35 %

No spring 76 J 59 J 218 J 414 J 33 %

Pump pressure [MPa]

Fig. 7 Pressure in the pressure line of the pump, during the stroke.

Fig. 8 Lowered pressure in the pressure line of the pump, during the stroke.

efficiency will be ca. 6 percentage points if the displacement is halved. 3.2.4 Intentional Pressure Drop (Hiccough) Previous results are simulated ensuring that pressure before the control valve remains near maximum. This is done by modeling large accumulator and volume in the pressure line. This means that pressure in the pressure line of the pump is staying near the maximum. In Fig. 6 and Fig. 7 the simulated pressures are shown

(non-return spring case). Next, only the relatively small hydraulic accumulator is placed before the control valve. After that pressure amplitude increases. The pressure must be kept in range that ensures that the movement of the actuator is unchanged. This could need the change of the controller parameters, too. After the accumulator resizing the pressure in the pump pressure line is shown in Fig. 8. The lower pressure demand causes that energy consumption from the pump decreases from 410 to 318 J (22% lower). Nearly similar decreasing (17%) is detected in return spring case. This is, however, achieved only if constant pressure pump is used. The variable displacement pump can typically adjust the flow/pressure demand within the 40-150 ms [5] and thus “ruin” the purposed pressure drop. The gas exchange valve can be lagging easily if environmental parameters are changed. The usage of rail (compressed volume) instead of accumulator should be slightly better, because of no back and forth flow is needed. However, the used simulation model did not find significant difference between them. Also, the previous simulations aren’t notifying the interference of the parallel cylinder or the actuators. The pressure in the pump pressure line can be significantly different in real, multi-cylinder case. The pressure needed in the actuator chamber during the stroke is presented in Fig. 9. However, this can be only achieved if actuator based pressure or flow control is applied [2]. Actuator pressure [MPa]

Pump pressure [MPa]

Fig. 6 Pressure in the pressure line of the control valve, during the stroke.

61

Actuator lift [m]

Pressure in P [MPa]


Fig. 9 Pressure in actuator upper chamber (return spring).

62


3.3 Energy Recovery The nominal required energy is defined by multiplying the force affecting to the actuator and the velocity of the actuator. Multiplying is done only at the moment when the acceleration is positive to the

Table 4 Energy in acceleration and deceleration phase of the actuator. Energy Opening Closing Recovery in opening Recovery in closing

Return spring (J) 118 36 9 47

No spring (J) 85 49 45 71

moving direction. Then the momentary product is integrated as shown in Eq. (3). This is the amount of

(3)

energy needed to make the open stroke. On the contrary, the deceleration part of the stroke can be assumed to be the energy which could be recovered. In this way the estimation of the recovered energy is better brought out. Because the frictions, the actuator will need some energy during the constant movement (and during the minor deceleration), but this matter is ignored in this study. The recovery partition also includes the energy needed to transfer fluid in return line. The energy needed to compress the return springs during the deceleration is subtracted from the opening recovery part. The maximum recovered energies during the stroke are presented in Table 4.

Fig. 10 Energy balance of the EHVA systems.

For comparison, if the gas exchange valve system would be displaced without any mechanical or viscous friction, the need of the energy would be around 60 J in opening direction (the return springs applied). In closing direction the springs can give all the energy needed. The energy stored in the two return spring is ca. 17 J. If the system is without return spring (mass only), the needed open/close energy is only 5/9 J. 3.4 Energy Distribution Energy distributions in both cases are presented in


63

Fig. 10. The efficiency of the hydraulic pump is now also included. The calculated actuator energy is the energy what is left after the wasted energy sources. Thus the recovered energy calculated could be even higher than residual actuator energy, because recovered energy includes part of the wasted energies.

control of the closed hydraulic circuit would have better efficiency than the throttle control of the open hydraulic circuit. There is up to 25% of energy which could be recovered, but this would need complex hydraulic systems and perhaps smart controlling system.

4. Conclusions

5. Future Work

The simulations showed that the largest energy loss (50-60%) is caused by the hydraulic control valve. This is predicted because all the deceleration energy is applied by pressure difference over the valve. Control valve with the better flow/pressure drop ratio would improve the situation. The hydraulic circuit has strong effect to the total amount of energy which EHVA system consumes. The usage of the constant return pressure and/or the return spring is lowering the hydraulic flow rate, and thus lowering the pressure/energy losses up to 25%. The system efficiency of the non-return spring system is also lower due to the multiple pressure losses over the hydraulic control valve. If the system flow demand is varying due to the rpm of the valve lift, variable displacement pump should be used. In that case the efficiency of the system will alter only amount of changes of the hydraulic pump efficiency. If the pressure could be adjusted to the demand of the actuator, significant energy saving up to 35% could be achieved. The variation of the pressure/flow demand is altering fast, and it could be hard to apply a simple and functional hydraulic system. The displacement

Different hydraulic circuit with different control valves are taken under investigation. The valve performance is essential to the moderate or good energy consumption. The actuator is slightly oversized and could apply more force than the system requires. Thus, the pressurized area could be decreased. The simulation model should be verified with the measurements.

References [1] [2]

[3]

[4] [5]

Y. Wang, Introduction to Engine Valvetrains, SAE International, Warrendale, 2007. J.H. Lumkes, Jr, W. Van Doorn IV, J. Donaldson, The design and simulation of a high force low power actuation system for camless engine, in: Proc. of Int. Mech. Eng. Congress and Exp., IMECE2005-81808, Orlando, Florida, 2005. M. Herranen, K. Huhtala, M. Vilenius, The electro-hydraulic valve actuation (EHVA) for medium speed diesel engines – development steps with simulations and measurements, in: SAE 2007 World Congress, Detroit, Michigan, 2007. Available online at: http://www.lmsintl.com/ D. Findeisen, F. Findeisen, Ölhydraulik, Handbuch für die hydrostatische Leistungsübertragung in der Fluidtechnik, (1994), Springer-Verlag, Berlin Heidelberg New York.

Call for Papers Dear author, This is Journal of Energy and Power Engineering (ISSN1934-8975), created in Dec. 2007. We are interested in various kinds of professional papers about energy and power. The Journal of Energy and Power Engineering, a monthly professional academic journal, is striving to provide the best platform for researchers and scholars worldwide to exchange their latest findings and results.

Current Columns Involve Thermal Science; Fluid Mechanics; Energy and Environment; Power System and Automation; Power Electronic; High Voltage and Pulse Power; Sustainable Energy. It is indexed by the well-known databases as follows: ★ Database of EBSCO, Massachusetts, USA ★ Database of Cambridge Science Abstracts (CSA), USA ★ Chinese Database of CEPS, American Federal Computer Library center (OCLC), USA ★ Ulrich's International Periodicals Directory, USA

Information for Authors 1. The manuscript should be original, and has not been published previously. Please don’t submit material that is currently being considered by another journal. 2. Manuscripts may be 3000-8000 words or longer if approved by the editor, including abstract, text, tables, footnotes, appendixes, and references. The title should not be exceeding 15 words, and abstract should not be exceeding 150 words. 3-8 keywords required. 3. The manuscript should be in MS Word format, submitted as an email attachment to our email address. 4. Author will receive 2 copies of the issue of the journal containing their article.

Submitting of Manuscripts All manuscripts submitted will be considered for publication. Please visit our website at www.davidpublishing.com for our automatic paper submission systems or as an email attachment to [email protected] and [email protected]. Should you have any questions or concerns, please feel free to contact us. Best regards Journal of Energy and Power Engineering David Publishing Company Tel: 1-847-281-9826; Fax: 1-847-281-9855

Journal of Energy and Power Engineering

Journal of Energy and Power Engineering

Suggest Documents

Journal of Energy and Power Engineering

Power and Energy Systems - College of Engineering and Computing ...

Knowledge Engineering Approach for Power - Journal of ...

POWER ENGINEERING GUIDE - Edition 7.0 - Siemens Energy

international journal of electrical power & energy

international journal of electrical power & energy ...

Turkish Journal of Engineering, Science and Technology Energy and ...

Electric Power and Energy Engineering: The First Century - IEEE Xplore

status of renewable energy in europe, us and ... - Engineering Journal

Electric Power and Energy Engineering: The First Century - IEEE Xplore

Journal of Power and Energy Systems - J-Stage

Journal of Power and Energy Systems - J-Stage

Energy Engineers, Part A: Journal of Power and

Journal of Power and Energy Systems - J-Stage

Energy Engineers, Part A: Journal of Power and

Energy Engineers, Part A: Journal of Power and ...

Journal of Power and Energy Systems - J-Stage

Journal of Power and Energy Systems - J-STAGE Journals

Energy Engineers, Part A: Journal of Power and

Energy Engineers, Part A: Journal of Power and

Journal of Power and Energy Systems - J-Stage

An Overview of Energy Sector and Wind Power ... - Journal Repository

Journal of Power and Energy Systems - J-Stage

Journal of Power and Energy Systems - J-Stage