Prevention of Transient Instability and Reactive Power ...

Available online at www.sciencedirect.com

ScienceDirect Aquatic Procedia 4 (2015) 1529 – 1536

INTERNATIONAL CONFERENCE ON WATER RESOURCES, COASTAL AND OCEAN ENGINEERING (ICWRCOE 2015)

Prevention of transient instability and reactive power mismatch in a Stand-alone wind-diesel-tidal hybrid system by an ANN based SVC Asit Mohanty*a Meera Viswavandyaa Sthitapragyan Mohantya a

CET Bhubaneswar,751003,India

Abstract This paper gives a novel idea of application of ANN based SVC controller for Reactive Power compensation in an isolated hybrid system and also discusses the improvement of stability in the hybrid system. For detailed analysis a small signal linear model of the hybrid wind- Diesel- tidal model is considered with different loading conditions. The reactive power compensation and stability analysis have been thoroughly analysed by a SVC Controller. A feed forward neural network with back propagation technique is designed to tune the parameters of SVC controller. Simulation result shows that the system parameters attend steady state value with lesser time and complexities. ©©2015 Published by Elsevier B.V. B.V. This is an open access article under the CC BY-NC-ND license 2015The TheAuthors. Authors. Published by Elsevier (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of ICWRCOE 2015. Peer-review under responsibility of organizing committee of ICWRCOE 2015 Keywords –Isolated Wind-Diesel; Tidal hybrid system; IG; DDPMG; SVC; ANN Controller

Nomenclature P,Q

IG/DDPMG

P Q SG SG

E

M

ΔQ ΔQ

K

, ΔE M

SVC COM

K K A, E, F

ΔV

Real Reactive power IG/DDPMG

K ,K a v

Exciter Gain, Gain Energy loop

Real Power, Reactive Power- SG

Ta , T , T r s ' X X d d

Exciter, rising, settling time const.

Electromagnetic energy and small change of Energy.(DFIG) Reactive generated by SVC

Δα

Reactive Power - Compensator.

ΔE

Gain Constants of Voltage Regulator, Exciter, Stabilizer. Incremental Change in Voltage

ΔE

η

q fd

Direct axis reactance of SG under steady state and transient Phase angle -Compensators Incremental Change in Internal Armature Voltage Change in the Voltage of the Exciter Performance Index

* Corresponding author. Tel.:+91- 9437920530; E-mail address:[email protected]

2214-241X © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of ICWRCOE 2015 doi:10.1016/j.aqpro.2015.02.198

1530

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

1. Introduction Standalone hybrid power systems are small power systems located at remote places to cater the local power demands of those particular places which are situated far away from the main grid. Generally two or more renewable sources are combined to form a hybrid system where shortage due to one source is compensated by the other.Bansal et al (2002). Wind Diesel systems are commonly used as hybrid system in which a diesel generator works as a backup with a wind turbine to provide power to remote places.Bansal et al.(2007). Normally Synchronous generators are preferred as Diesel Generator and SCIG/DFIG/PMIG are preferred in Wind Turbine for an better performance and for their rugged characteristics .Devraj et al(2011) and Hasemi et al(2013).In a wind– diesel-tidal hybrid system consisting of both synchronous generator and Induction generator, the Induction generator needs reactive power and is mostly provided by the Synchrous generator.juardo et al(2002) and Jaybharati et al(2007) But the supplied reactive power is not sufficient and it creates gap between the demand and supply of reactive power. This gap leads to problems like voltage fluctuation and instability .Generally capacitor banks are utilised to compensate reactive power in the power system. As renewable like wind is unpredictable and loads are constantly changing fixed capacitors cannot meet the challenge to compensate reactive power.Karami et al(2011).The challenges of power quality issues like voltage instability and reactive power compensation are generally met by the use of FACTS(Flexible AC Transmission System) devices.Kaldellis et al(2011) and Karemi et al(2009). SVC, STACOM and other FACTS devices are used to compensate the reactive power of the power system. They are extensively used for voltage and angle stability studies in power system.Mohanty et al(2014) Management of reactive power has become an important aspect of hybrid power system and absence of reactive power forces the system to go through a wide voltage variations and lots of fluctuations. Many a times SVC has been projected as a reactive power compensating device in several literatures. This work proposes an ANN based SVC Controller for reactive power management and stability enhancement in a standalone wind diesel tidal hybrid system.Bansal et al (2007) The proposed controller helps in tuning the gains of the PI Controller to enhance the transient stability and reactive power compensating capability of the system.Numerous problems are easily solved with the application of ANN controller in power system. The working of ANN is similar to working of biological nervous systems. ANN consists of a large number of interconnected processing elements (neurons) working together to solve a problem. It is designed for a specific application through learning process which adapts the synaptic connections of the neurons. The main merits of ANN are the relationship between the input and output data s for an unknown relationship or complex function. The gains of the PI controller depend upon the type of reactive power load for optimum performance. Due to the variable nature of the load, the PI gains setting of SVC are adjusted after proper tuning.Padiyar et al (1991) and Padiyar et al (2008). The paper focuses the ANN based approach to tune the PI gains of the SVC controller over a wide range of load characteristics. For the simulation the multi-layer feed-forward ANN tool box of MA TLAB with the error back-propagation training method is used. Riedmiller et al (1993). The dynamic responses of the hybrid system are shown for 5 % step increase in load reactive power with and without 5% step increase in input wind power. This paper discusses a small signal model of wind-diesel-tidal system for analysis of transient stability and reactive power compensation with introduction of ANN based SVC controller for 5% load change. Section II describes the whole system and the detailed mathematical model of it. Detailed work with application of of ANN controller for tuning the gains of PI controller is discussed in section III. Finally the simulation results with detailed descriptions are represented in section IV and conclusion part in Section V. 2. System Configuration and its Mathematical Modelling The said structure of standalone wind-diesel-tidal hybrid system basically consists of Induction generator (IG) based wind turbine, DDPMSG based tidal turbine with Synchronous generator (SG) based diesel engine as backup. The block diagram is clearly depicted in Fig. 1.For a system like this, the active power is provided by the Induction generator and Synchronous generator. But for reactive power need of the Induction generator and Load an SVC with ANN controller is provided. SVCs are considered better reactive power compensators in comparison to other FACTS devices. The hybrid system parameters are mentioned in Table.

1531

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

Reactive power

Wind System Wind

V ‘0

IG IG

Reactive power

Tidal System DDPMG

Vt

SVC Reactive Power

1 1+sTA

Reactiv e Power

SG

SG Fuel

Regulator

Reactive power

6

KA

Feedback

Consume r Load

S E = f(E fd )

V R M AX

V rref

1+STA

S Signal

Exciter

Diesel Generator Set

Saturation

sK F

³ V R M IN

1 + sT F

1 KE +sTE

Exciter

BusBar

Fig.1.(a) Wind-diesel-tidal hybrid Power system with SVC ;(b)Excitation System 1.

'Vref IG

Load

PI Controller KI KP  s

Kl

'Q L

'QIG

Ke

'V(s)

DDPMG

SVC

KV 1 sT V

Kk

Kj 'QCOM

'QDDPMG 'QSG

1 0.75 s 1

Kc

Kb

Kd Ka Saturation Function

PI Controller

'Vref

KI KP  s

sK F 1 sT F

KA 1 sTA

AVR

SF KE 1 sTE

E fd

Fig.2. Small signal transfer function model of Wind-diesel-tidal hybrid system with SVC

The balanced reactive power equation of (SG, SVC, IG, DDPMG and LOAD) is expressed as (1)

ΔQ + ΔQ = ΔQ + ΔQ + ΔQ SG SVC L IG DDPMG

Due to small change in reactive power load ∆Q L, the system terminal voltage varies which affects the reactive power requirement of other members of the system. The final reactive balance equation is equal to ΔQ +ΔQ -ΔQ -ΔQ -ΔQ 0 and the value deviates the system output voltage. The System Model SG SVC L IG DDPMG Equation is governed by the transfer function equation which is given below [14] ΔV(S) =[ΔQ (S)+ ΔQ (2) (S)-ΔQ (S)-ΔQ (S)-ΔQ (S)] SG COM L IG DDPMG 2.1. Wind Turbine (Induction Generator)

The wind turbine uses a squirrel cage Induction generator. The real and reactive power delivered by the induction generator are given by RY 2 PIG = V 2 R2 +X Y eq

ª

and QIG = «

Xeq

2 «¬ R Y

+ X2 eq

+

1 º 2 »V XM »¼

The reactive power absorbed by the Induction Generator is given by

(3) ­ ½ ° Xeq ° 2 QIG = ® ¾V 2 2 ° RY +Xeq ° ¯ ¿

2.2. Tidal Turbine (Direct Drive Permanent Magnet Generator) Tidal energy is one of the prominent renewable technologies for generation of electric power. The Tidal turbine mainly uses uses Doubly Fed Induction Generators (DFIG) or Direct Drive Permanent Magnet Synchronous Generators (DDPMSG).The tidal current is a new source of energy and is very effective as it uses the same technology like wind energy system. The use of tidal currents as a new source of energy is a very effective source as

1532

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

it relies on the same technologies used in wind turbines and it is a predictable source of energy. The tidal power is captured using tidal turbines. Tidal powers are much more predictable than wind and solar energies. As we know Tidal stream and tidal barrage are two main classifications for generating energy from the tides.The tidal stream uses the kinetic energy of moving water and tidal barrage uses the potential energy in height respectively. The energy captured from tidal stream 1 2 PT = ρ.A.Cpt(λ,β).Vt 2

Where U is the density of water, A is the swept turbine area, C pt is the turbine power

coefficient and Vt is the velocity of water flow.

Vs

V DC _ ref

IS R  Z L S S

V

iDg _ ref

k k + i3 p3 s

iD g

D C

V*Dg

k k + i5 p5 s

VDg

Xc

Z \

V i _ ref

V

k k + i4 p4 s

iQ g

Xc k k + i6 p6 s

iQg _ ref

V Qg V*Qg

i

Fig. 3.(a) Equivalent circuit of DDPMSG ; (b)Grid side converter block diagram.

2.3. Diesel Engine (Synchronous Generator) For the Diesel system, the modified synchronous generator equation is QSG = Vcosδ

and for incremental change ΔQSG = ' ' + X dΔE q

(E'qVcosδ-V2) (Transient condition) X'd

E'qcosδ-2V X'dΔV

(4)

Taking Laplace we get ΔQSG(s) = Ka ΔE'q(s) + KbΔV(s) Where Ka= Vcosδ /X'd and Kb= (E'qcosδ -2V)/X'd 2.4. Static Var Compensator (svc) The main function of SVC is to provide reactive power required by the load and the Induction generators which can not be met by the synchronous generator alone. It is a high speed device whose function is to compensate the reactive power and meet the voltage stability problem. The variable Inductive susceptance BL is a function of the Thyristor firing angle D and is equal to BL

23  2D  sin 2D 3 , d D d 3 , X i = reactance of the fixed Inductor Xi 2

of the SVC. When SVC compensates the reactive power QSVC =V²BSVC and the Laplace Transform for a small perturbation [16] (5) ΔQ (s)= K ΔV(s)+ K ΔB (s) svc c d svc 'V (s)

Thyristor firing delay D 0  'D d 3

'Vref (s)

KP 

Ki s

'D (s)

Phase sequence delay 'B 'SVC ( s) 'B (s)

KD 1 sTD

3 d D 0  'D 2 Fig. 4. Small signal model of Static Var Controller

1 1 sTd

SVC

x 1 x 2 x n Input Layer

D1 D2

Transfer Function Output

D3

¦

D4

Activation Function

Threshold Output Layer

Dn Hidden Layer

Fig. 5. Artificial Neural Network

3. Artificial Neural Network ANN is basically based upon the neural structure of the brain. It tries to imitate the functioning of the brain. As we

1533

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

know that brain stores information and ANN provides a new field of computing which involves the creation of massive parallel networks to solve specific problems. As neurons provide the ability to remember and apply the previous experiences, ANN works on a similar pattern to achieve high computational rates due to massive parallelism fault tolerance capability. The controller uses back propagation algorithm for the training process. 3.1. Back Propagation Algorithm This algorithm is the most popular and widely accepted algorithm. First step-Initialization--All the weights, threshold levels of the network are randomly distributed in a small range. Second step-Activation--Here the actual outputs of the neuron in the hidden layers are calculated. ª n º sigmoid « ¦ xi ( p ).wij ( p ) Ti » «¬i 1 »¼

y ( p) j

Where n is the number of inputs of the neurons j in the hidden layer.

So the actual outputs of the neuron in the outer layer

y ( p) k

ªm º sigmoid « ¦ x jk ( p ). wik ( p )T k » ¬«i 1 ¼»

Third step- Training of weights-The error gradient for the neurons in the output layer is calculated. y ( p) ª¬1 yk ( p)º¼ .e ( p) k k

G k ( p) 'w

jk

( p)

, ek ( p)

y ( p)  y ( p) And the weight corrections are dk k

D . y j ( p )G k ( p ) , then the weights at the output neurons are updated as 'w jk ( p  1)

w

jk

( p )  'w

jk

( p)

The error gradient of the hidden layer neurons can be calculated G j ( p)

l y ( p ) ª¬1 y j ( p ) º¼ ¦ G k ( p ) w ( p ) j jk k 1

updated the weights at the hidden neuron w

and the weight corrections are 'w jk

( p  1)

w

jk

( p )  'w

jk

jk

( p)

D x ( p )G ( p ) , When i

j

( p)

Fourth step-Iteration-An increase of iteration p by one step and the process should be repeated until the selected error Criterion is satisfied. 3.2. Training of parameters by ANN The value of the reactive power load voltage characteristics ( nq ) is used as input of the ANN and the outputs are the proportional and integral gains k p and ki of SVC controller. The ANN uses the normalised values of nq as input and produces output in normalised way and is converted to actual one. Determination of weights is known as training of the learning process. An input output pairing is first prepared prior to the conducting of training process. Based on the loading conditions the set is first developed by calculating the desired PI controller gains. A typical range between 0.0 to 4.0, taking the exponent of load voltage characteristics is prepared to cover all typical loads of the power system. The network is trained repeatedly till an optimum agreement between predicted gains and actual gains is achieved. The network is again tested to predict the actual gain settings of the load model. The performance index

K is based on integral square error (ISE) and is equal to k and k are determined. i p

performance index optimised values of min

constraints. K p

dK

p

dK

max p

and

typical values of nq , 1.25 and 3.25. Table1. Optimum gain settings of SVC

K

K

min max and by minimising the dK dK i i i

'V is

the voltage deviation, subject to the

min max .The testing of the networks is performed for two dK dK i i i

1534

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

n Optimum gain settings of SVC for different values of exponent of reactive-load voltage characteristics q kp

0.5 35

1 38

1.5 41

2 45

2.5 47

3 49

3.5 50

4 54

ki

4985

5100

5405

5600

5807

6008

6234

6458

4. Simulation Results From the MATLAB based simulation which has been carried out taking an ANN based SVC Controller for the wind diesel tidal hybrid power system the settling points and peak overshoots of different parameters of the hybrid system are noted..The ANN controller has been utilised for compensating reactive power and voltage stability of the hybrid system with a step load change of 5%. The variation of all the system parameters such as small change in reactive power of Synchronous generator, Induction generator, DDPMSG, Reactive Power Change of SVC, Variation in firing angle, Variation in terminal voltage, Variation in field excitation, Variation in armature voltage, and change of armature voltage under transient, etc., as shown in Fig. 6 are studied for the above mentioned disturbance using traditional PI Controller and the ANN controller.

1.5

4

No Controller SVC(PI) SVC(ANN+PI)

Angle 2

1

QSvc

No Controller SVC(PI) SVC(ANN+PI)

0.5

0 -2 0

0.1

0.2

0.3

0.4

0.5

Time(sec)

0.6

0.05

0.7

0 0

0.1

0.1

-0.05 0.2

0.3

0.4

0.5

0.6

-0.1 0

0.7

Time(sec) 0.05

0.01

0

-0.1

0

0.1

0.2

0.3

0.4 Time(sec)

0.5

0.7

No controller SVC(PI) SVC(ANN+PI)

DelDDPMG 0.1

0.2

0.3

0.4

Time(sec)

0.5

0.6

0.7

DelEq' No controller SVC(PI) SVC(ANN+PI)

0 -0.005 0.6

-0.01 0

0.7

x 10 0.05

0.1

0.2

0.3

0.4

Time(sec)

0.5

0.6

0.7

-3

5

0

0

0.1

0

No controller SVC(PI) SVC(ANN+PI)

DelV

-0.05 -0.1

0.6

0.005

No controller SVC(PI) SVC(ANN+PI)

DelEq

-0.05

0.5

0

DelEfd 0

0.4

Time(sec)

0

-0.1

0.3

0.05

No controller SVC(PI) SVC(ANN+PI)

-0.05

0.2

0.2

0.3

0.4

Time(sec)

0.5

-5

0.6

0.7

-10 0

No controller SVC(PI) SVC(ANN+PI)

DelQIG 0.1

0.2

0.3

0.4

0.5

0.6

Time(sec)

Fig. 6. (a-h) Output of Wind Diesel tidal hybrid system using SVC for 5 % load change during transient condition (comparison with ANN).

0.7

1535

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

0.06 0.05 0.04

No contoller

0.03 0.02

PI

0.01

ANN

0

DelV

DelDDPMG

DelEq'

Del efd

0.8 0.6

no control

0.4

PI

0.2

ANN

0

DelV

DelQSG

DelQIG

DelDDPMG

DelEq'

Del angle

Del SVC

Del efd

Fig. 7.(a) Maximum deviations; (b) Settling time of parameters with time with 5% load change. Table 2. Comparison between two controllers for 5% load change PI Controller

DelV Del SVC Del QSG Del QIG Del DDPMG Del Eq’ Del Alpha Del Efd

ANN Controller

Maximum Deviation

Settling Time

Maximum Deviation

Settling Time

0.05 1.5 0.52 0.0052 0.048 0.01 4.1 0.049

0.5 0.6 0.5 0.42 0.4 0.61 0.55 0.64

0.048 1.48 0.515 0.0051 0.047 0.009 3.4 0.048

0.475 0.58 0.48 0.4 0.38 0.6 0.52 0.6

The settling time and peak overshoot in the case of ANN controller is observed to be less as comparison to traditional PI Controller. Simulation results clearly show the output of ANN Controller is better than the traditional PI Controller. 5. Conclusion This paper gives a novel solution for the transient stability analysis and reactive power compensation in the winddiesel-tidal hybrid system with the incorporation of ANN based SVC controller. Results show the Variation of the system parameters with variation of load and it is quite evident that the reactive Power compensation of the model has been achieved .The compensation in case of back propagation based ANN controlled system is better than the conventional system. The proposed ANN controller based hybrid system shows better results in settling time and overshoot. The ANN Controller shows effective improvement and changes in the output by tuning the values of KP and Ki than the conventional PI controller and thereby helps in improving the stability of the system. Appendix A. Parameters of wind-diesel hybrid system System Parameter Wind Capacity (Kilowatt) Diesel Capacity Tidal Capacity Load Capacity Base Power in KVA SYNCHRONOUS GENERATOR PSG, KW 0.4 QSG in KW 0.2 Eq( pu) 1.113

INDUCTION GENERATOR PIG,pu in Kilowatts 0.6 QIG,pu in KVAR 0.189 PIN,pu in Kilowatts 0.75

WindDieselTidal syst 150 150 150 400 400 DDPMG Lss 0.17pu Vref 1 pu Rs 0.01 pu

1536

Asit Mohanty et al. / Aquatic Procedia 4 (2015) 1529 – 1536

Eq’(pu) 0.96 V(pu) 1.0 Xd,(pu) 1.0 T’do,s 5 LOAD PL(pu) in KW 1.0 QL(pu) in KVAR 0.75 α in Radian 2.44

r1=r2(pu) X1=X2(pu)

0.19 0.56

Ψ

1.1pu

References Bansal, R.C., December (2002).Automatic Reactive Power Control of Autonomous Hybrid Power System, PhD Thesis, Centre for Energy Studies, IIT Delhi. Bansal, R.C., Bhatti, T.S., Kumar, V. (2007).Reactive Power Control of Autonomous Wind Diesel hybrid Power System using ANN, Proceeding of the IEEE Power Engineering Conference Devaraj, D., Preetha Roselyn, J., (2011). On-line voltage stability assessment using radial basis function network model with reduced input features. Int J Electr Power Energy Syst.pp1550–51. Hashemi, S., Aghamohammadi, MR., (2013). Wavelet based feature extraction of voltage profile for online voltage stability assessment using RBF neural network. Int J Electr Power Energy Syst.pp86–94. Hagan MT, Menhaj M., (1994). Training feed forward networks with the Marquardt algorithm. IEEE Trans Neural Networks.pp.989–93. Hingorani, N.G., Gyugyi,L.,(2000).Understanding FACTS, Concepts and Technology of Flexible AC Transmission System, NewYork, IEEEPowerEngineering Society. Juardo, Francisco., Saenz José, R. (2002).Neuro Fuzzy Control for Autonomous Wind Diesel System using Biomass, Renewable Energy Jayabarathi,R.a., and Devarajan,N.b.(2007). ANN based DSPIC controller for reactive power compensation, American journal of applied science, pp.508-515. Karami, A., (2011). Power system transient stability margin estimation using neural networks. Int J Electr Power Energy Syst, pp983–91. Kaldellis, Jet al. (1999).Autonomous energy system for Remote Island based on renewable energy sources, in proceeding of EWEC, Nice. Mohanty, Asit., Viswavandya Meera., Ray Prakash K., Patra Sandipan.( 2014).Stability analysis and reactive power compensation issue in a micro grid with a DFIG based WECS, Electrical Power and Energy System. Murthy, S.S., Malik, O.P., and Tondon, A.K. (1982).Analysis of Self Excited Induction Generator, IEE Proceeding, vol 129. Padiyar, K.R., (2008).FACTS Controllering in Power Transmission system and Distribution, NewAge International Publishers. Padiyar, K. R., Verma ,R.K.(1991).Damping Torque Analysis of Static VAR System controller,IEEE Transaction of Power System, PP458-465. Riedmiller, M., Braum, H. (1993). A direct adaptive method for faster back-propagation learning: the RPROP algorithm. IEEE Int Conf Neural Networks.