March 26-29
Wind Power Integration in Hybrid Power System Active Energy Management M. A. Tankari, M.B.Camara, B. Dakyo, C. Nichita GREAH Laboratory, University of Le Havre, Le Havre, France, 25 rue Philippe Lebon, BP 540, 76058 LE HAVRE Cedex France E-mail:
[email protected] E-mail:
[email protected] Abstract: The wind speed is assumed to be a sum of slow and turbulence components containing the total energy. The caption of such large frequency band of energy will increase the profitability of a hybrid wind-diesel power source. Indeed, to absorb all fluctuations of the wind generator induced current, it is necessary to oversize diesel generator, or to connect storage devices for medium and high frequencies components absorption. Inserting a flywheel to absorb the high frequencies would increase significantly the battery lifetime. An original strategy presented in this paper is based on the frequency share of diesel and storage behaviour ensuring a better lifetime of these sources and meeting the demand of the load. The effectiveness of this strategy lies in its ability to control the frequency bands allocated to the diesel engine and to extend the batteries lifetime by individual charge/discharge cycles moderation. The aim of this paper is to analyze the degree of selectivity and the configurations of filters through a specific study of the spectral density of signals. The batteries lifetime is then evaluated using the rainflow cycle counting algorithm for effectiveness comparison. Keywords: Hybrid energy system, Energy storage, battery, wind turbine, flywheel, power spectral density, diesel engine. 1. Introduction 2. Wind speed model Wind diesel hybrid system is subject to disturbances due to wind speed fluctuations. Long life and low capital cost of diesel engine is guaranteed when storages devices are used. According to energy storage issue, the properties of available technologies can be spread on dual energy-power definition related to time horizon of effectiveness. This paper deals an energy transfer management strategy in hybrid system including wind turbine, diesel engine, flywheel, battery, capacitor, and emulated load management. In first case, the importance of taking into account of wind speed high fluctuations is introduced. In second case, the hybrid system management is presented through first and second order low pass filters studies. Finally, some simulations results are presented and analyzed.
1
The wind speed spectrum shown in Fig. 1 was constructed by Van Der Hoven (1957) [1] from long- and mean-term records at Brookhaven, New York. Few days’ variations (synoptic), variations with the time of day (diurnal) and shorter time-scales of minutes down to seconds or less wind-speed variations (turbulence) are observed. The phenomenon called ‘spectral gap’ occurs between the diurnal and turbulent peaks. It shows that the synoptic and diurnal variations can be treated as quite distinct from the higherfrequency fluctuations of turbulence. There is very little energy in the spectrum in the region between 2 h and 10 min. These turbulent fluctuations then have a zero mean when averaged over about 10 min to 2h. . This phenomenon is not observed in data analysed in [2] that show a smooth behaviour through de concerned region. The ‘spectral gap’ may
occur when measures at various altitudes or different sampling periods are combined [2]. One of the important information from Fig. 1, which is correlated by [2], is that high frequency fluctuations contain considerable power and it can have a very significant effect on the design and performance of the individual wind turbines, as well as on the quality of power delivered to the network and to the consumers. In [2], acquired data of 104 turbines located at 4 sites were studied. It is shown that wind speed frequency bandwidth is from 2.5*10-7 Hz to 1.4*10-4 Hz when measurements are carried out with a sampling period of 1 hour on 366 days, while it lies between 9.2 * 10-9 Hz and 0.5 Hz for a sample of 1s on 10 days. The significant part of the generator power fluctuations is located in the low (below 0.01 Hz) and medium (between 0.01 and 1 Hz) frequency regions. Most of the high frequency fluctuations (above 1 Hz) are effectively damped out by the large inertia of turbine generator; their magnitudes are insignificant [3].
are required to match medium and high fluctuations. The lifetime of the battery would be multiplied by four when a flywheel is used to absorb fluctuations in high frequencies instead of using only the battery [6]. The hybrid system presented in Fig. 2 is DC bus linked.
Fig. 2. Hybrid system with flywheel
The energy transfer between different sources and load is based on filtering wind power to generate the current reference of storage devices. The batteries allocated frequency band is 1.3mHz to 20 mHz. The flywheel absorbs high frequency currents (over 20mHz) and the diesel engine is used to regulate the DC bus voltage continuously (below 1.3mHz). 4. Filtring Principle
Fig. 1. Wind Spectrum Farm Brookhaven Based on Work by van der Hoven (1957)
So, the wind speed v(t) can be expressed as the sum of the harmonics characterised by the magnitudes Ai, the pulsation wi and the phase ϕi generated randomly [4][5]. v(t) = vl (t) + vt (t) =
2
Nl
2
N
∑ A cos(wt +ϕ ) + π ∑ A cos(wt +ϕ ) π i =0
i
i
i
i
i
i
Nt
(1) Nl are Samples for the slow component vl(t) and N-Nt are Samples for the component of turbulence vt(t). 3. Hybrid system management The energy capacity of diesel generator operating in slower fluctuations would be double sized to fill in fast fluctuations with amplitudes of 1% of the maximum fluctuation [2]. The fast-ramp-rate energy storage systems such as flywheel, batteries or supercapacitors
2
The hybrid system management is based on filtering of wind turbine current by two lowpass filters. The optimal current Iwind_opt provided by wind turbine to the DC bus is measured and filtered through the low pass filter HLPF1(p) whose output is subtracted from Iwind_opt to get the reference current IFly_ref of the flywheel regulator (Fig.3). The filter HLPF1(p)cut-off frequency fc1 corresponds to the battery allowed maximum frequency (20mHz in this case). As shown by fig.4, resulting signal from subtraction of measured flywheel actual current IFly from wind turbine current Iwind_opt is filtered by HLPF2(p). The filter output is subtracted from its entrance, to generate the battery reference current Ibat_ref. The filter HLPF2(p)cut-off frequency fc2 corresponds to the battery allowed minimum frequency (1.3mHz in this case). Using this strategy, the diesel generator current is expected to be very low frequency.
The effectiveness of this method relies on the ability of the flywheel to absorb high fluctuations currents. That is why we will focus on the control of the flywheel in the next section.
Bode Diagram 20
Cut-off Frequency : 20mHz Magnitude (dB)
0 1st Order
-20
2nd Order, Q=0.05 -40
2nd Order, Q=0.5 2nd Order, Q=0.707
-60
2nd Order, Q=5 2nd Order, Q=20
-80 0
Phase (deg)
-45 -90 -135 -180
Fig. 3: Flywheel reference current
-5
10
-4
10
-3
10
10
-2
10
-1
0
1
10
10
Frequency (rad/sec)
Fig. 6: 1er and 2nd Low Pass Filters (Fc=20mHz)
5. Simulations results The wind diesel hybrid system is simulated in a runtime of 7000s in various configurations based on low pass filters of first and second order. The effects of three characteristics quality factors (0.5, 0707, and 20) of secondorder filters are analyzed and compared to those of the first order filter of the same frequency. The profile of the battery current generated in the case of a 2nd order filter quality factors of 0.5 is quite close to that with a quality factor of 0707 (Fig.7). When Q varies from 0.5 to 20, it can be observed that current profile of the battery becomes more fluctuant (fig.7 and fig.8), which increases the number of charges / discharges partial cycles and reduce the battery’s lifetime.
Fig. 4: Battery reference current
A First order Low pass filter The first order low pass filter transfer function is expressed by (2).With fc=1/(2.π.τc) the filter cut-off frequency and τ c the time constant [7] [8].
1 H LPF ( p )= 1+ pτ c
(2)
B Second order Low pass filter The second-order filters attenuate more the signal’s magnitude beyond the cutoff frequency. Second order low pass filter transfer function is expressed by (3).
τc p
Q
+ (τ c p )
30
(3) 2
20 Current (A)
1+
Low Pass Second Order Q=0.707 Low Pass Second Order Q=0.5
1
H LPF ( p ) =
Battery Currents 40
With Q the quality factor which effects are illustrated by fig. 5 and fig.5; The signal near cut-off frequency, is amplified when Q is higher than 0.707 and is reduced when Q is lower than 0.5.
10
0
-10
1400
1600
1800 Times (s)
2000
2200
2400
Fig. 7 Zoom of instantaneous battery currents
Bode Diagram 20
Cut-off Frequency : 1.3mHz Magnitude (dB)
0
Fig. 9 and 10 suggest that it is more efficient to use a filter of first order than a 2nd order filter with quality factor equal to 20. As shown by the current spectral density (fig. 10) the quality factor of 20 led to a considerable amplification of the battery current value beyond the cutoff frequency.
1st Order
-20
2nd Order, Q=0.05 -40
2nd Order, Q=0.5 2nd Order, Q=0.707
-60
2nd Order, Q=5 2nd Order, Q=20
-80 0
Phase (deg)
-45 -90 -135 -180 -5
-4
10
10
-3
10
-2
10
-1
10
0
10
Frequency (rad/sec)
Fig. 5: 1er and 2nd Low Pass Filters (Fc=1.3mHz)
3
1
10
Battery Currents
Battery Currents spectrum 100
30
90
20
80 70 Magnitude
10 Current (A)
1st order LPF 2nd order LPF(Q=0.5)
0 -10 -20
60 50 40 30
-30
20
Low Pass Second Order Q=20 Low Pass Second Order Q=0.707
-40 1200
1300
1400
1500
10
1600 1700 Times (s)
1800
1900
2000
2100
0 -4 10
-3
10
-2
10
Fig. 8 Zoom of instantaneous battery currents
-1
10 Frequency (Hz)
0
10
1
10
2
10
Fig. 11: Battery currents spectrum Battery Currents
Battery Currents spectrum 100
60 Low Pass Second Order Q=20
2nd order LPF(Q=0.707) 2nd order LPF(Q=0.5)
90
Low Pass First Order
40
80
30
70
20
60
Magnitude
Current (A)
50
10 0 -10
50 40 30
1900
1950
2000 2050 Times (s)
2100
20
2150
10
Fig. 9 Zoom of instantaneous battery currents
0 -4 10
10
-3
-2
10
-1
10 Frequency (Hz)
0
10
1
10
2
10
Battery Currents spectrum 200
Fig. 12: Battery currents spectrum
1st order LPF 2nd order LPF(Q=20)
160
Magnitude
140 120 100 80 60 40 20 0 -4 10
-3
10
-2
10
-1
10 Frequency (Hz)
0
10
1
10
2
10
Fig. 10: Battery currents spectrum
Fig. 11 to 13 show good attenuation at frequencies above 20mHz in the case of first order filters and second-order filters with factor of quality between 0.5 and 0.707 have similar spectra at frequencies above 1.3mHz and good attenuation at frequencies above 20mHz. Against by the magnitude spectrum is amplified at frequencies below 1.3mHz by second-order filters. The amplification increases with the decrease of the quality factor from 0.707 to 0.5 (fig.11 and 12). The flywheel power spectrums (fig. 13 and fig.14) show the complexity to analyze the filtering effect because of the high density and low amplitudes of the wind turbine fluctuating currents. It is observed that the magnitudes of the spectrums frequencies below 20mHz are higher than those of the flywheel allocated frequency band.
Indeed, it is finding that the magnitudes of the battery currents are greatly larger and less dense than those of flywheel currents at frequencies below 20mHz. Beyond this frequency, the flywheel currents are larger and denser. The filtering principle has therefore effect to allocate to the flywheel the densest and lowest magnitudes signals along the frequency band of wind currents, thereby reducing the number of battery partial cycles (fig.15). The flywheel spectrum does not show much difference in the case of the first filter and second order filter with factor equal to 0.5 and 0707. Exchanged currents between wind turbine, diesel engine and load with first order filter control are illustrated by fig.16. Flywheel Currents spectrum 30 1st order LPF 2nd order LPF(Q=0.5) 25
20 Magnitude
180
15
10
5
0 -4 10
-3
10
-2
10
-1
10 Frequency (Hz)
0
10
1
10
Fig. 13: Flywheel currents spectrum
4
2
10
Flywheel Currents spectrum
complete discharge and recharge of a fully battery [9][10]. From data provided by the batteries manufacturers, the approximated expression of the curve of numbers of cycles depending on the amplitude of discharge current is calculated (eq.4) [10]
30 2nd order LPF(Q=0.707) 2nd order LPF(Q=0.5) 25
15
10
5
C F = a1 + a2 e − a3 R + a4 e − a5 R -3
-2
10
-1
10
0
10 Frequency (Hz)
10
1
10
10
Fig. 14: Flywheel currents spectrum Flywheel and Battery Currents 60 Flyw heel (2nd Order LPF Q=0.5)
50
Battery (2nd Order LPF Q=0.5)
40
Current (A)
30 20 10 0 -10 -20 -30 -40
0
1000
2000
3000 4000 Times (s)
5000
6000
7000
Fig. 15: Flywheel and Battery currents Currents 700 Wint turbine current Load current Diesel engine current
600
Magnitude (A)
400
300
200
100
1000
2000
3000 4000 Times (s)
Avec, CF : cycles to failure ai : constants to determine R : cycles magnitude (fraction of discharge depth); The lifetime is then calculated using the Palmgren-Miner rule [10] [12] considering that the fraction of life consumed during a given cycle is 1/CF. When the sum of the numbers of cycles multiplied by their corresponding fraction exceeds 1, it is considered that the battery is dead and must be replaced. Considering Ni cycles of 20 fractions of discharge depths classified from 0.05 to 1. (i.e. 5% to 100%) the total damage, D, over a given time, is expressed by (5). 20 1 (5) D = ∑ Ni CF ,i
i =1
500
0 0
5000
6000
7000
Fig. 16: Wind turbine, Diesel and Load currents
As shown by simulation results, the battery charge state typically does not follow a regular pattern of cycle and spectral profiles analysis are not sufficient to conclude on the effectiveness of a filter in relation to others. The study will be completed by calculating the lifetime of the battery. Method used in this paper supposes that the number of cycles a battery can tolerate is only function of the discharge current depths.
With CF,i the number of cycles to failure according with class i. Assuming that the battery returns to its initial state of charge after two hours of the hybrid system’s operation, simulation results can be extrapolated to the years to determine the battery life. The software KiBaM elaborated by RERL [13] is used to determine the numbers of cycles according to their depths for each type of filter (fig. 17 to 20). Table 1 presents the OPzS battery lifetime with different filter types. A second order filter with quality factor equal to 0.5 tends to ensure better life of the battery. Battery Current (2nd Order LPF Q=0.5)
Battery Current (2nd Order LPF Q=0.707)
6
6
5
5
4
3
2
1
The counting algorithm called rainflow cycle counting is applied to the time series of the battery currents to find the individual charges/discharges cycles, which are then stored into classes (usually 20 classes are used) of equal size. The classes correspond to different depths of discharge, and the last class corresponds to the 5
(4)
2
No. of Cycles (cycles)
0 -4 10
No. of Cycles (cycles)
Magnitude
20
0
4
3
2
1
0
0.01
0.02
0.03
0.04 0.05 0.06 Cycle Depth
0.07
0.08
0.09
0.1
0
0
0.01
0.02
0.03
0.04 0.05 0.06 Cycle Depth
0.07
0.08
0.09
0.1
Fig.17 Number of cycles Fig.18 Number of cycles
Battery Current (2nd Order LPF Q=20) 12
5
10
No. of Cycles (cycles)
No. of Cycles (cycles)
Battery Current (1st Order LPF) 6
4
3
2
1
0
8
6
4
2
0
0.01
0.02
0.03
0.04 0.05 0.06 Cycle Depth
0.07
0.08
0.09
0.1
0
0
0.01
0.02
0.03
0.04 0.05 0.06 Cycle Depth
0.07
0.08
0.09
0.1
Fig.19 Number of cycles Fig.20 Number of cycles Table 1. Battery Lifetime’s estimation for different filter types. Low 1st 2nd 2nd 2nd Pass Order Order Order Order Q=0.5 Q=0.707 Q=20 Filter Lifetime (years)
1.8
3.1
2.2
1
6. Conclusion In this paper, it is showed that wind power due to high frequencies is considerable and that taking it into account increases wind-diesel hybrid system performance. As against, while the diesel should compensate all wind current fluctuations it should be double sized compared to where it should carry only lowfrequency wind currents. In addition, a diesel operating at high dynamics lifetime is considerably reduced. Big savings will be made so by using storage devices to mitigate the wind currents of medium and high frequency but much lower amplitude compared to the low frequencies current. The original strategy proposed is based on frequency allocation of diesel, battery and flywheel missions by cascading two first order filters. Second order filters quality factors and first order filters effects are studied. Spectral density analysis is completed by battery lifetime estimation to compare the effectiveness of each filter configurations. Simulations results highlight the interest to operate battery with second order filter. References [1] T. Burton, D. Sharpe, E. Bossanyi, “Wind Energy Handbook”, John Wiley & Sons, Ltd, 2001 [2] J. Apt, “The spectrum of power from wind turbines”, Journal of Power Sources 169 (2007) 369-374; [3] W. Li, G. Joós, and C. Abbey, “Attenuation of Wind Power Fluctuations in Wind Turbine Generators using a DC Bus Capacitor Based Filtering Control 6
Scheme” Industry Applications Conference, 2006. 41st IAS Annual Meeting. Conference Record of the 2006. IEEE Vol. 1, Oct. 2006 Page(s):216 - 221 [4] C. Nichita, D. Luca, B. Dakyo and E. Ceanga, “Large band simulation of the wind speed for real time wind turbine simulators”, IEEE Transactions on energy conversion, Vol 17, No. 4, Dec 2002; [5] B. G. Rawn, P. W. Lehn, M. Maggiore, “Control Methodology to Mitigate the Grid Impact of Wind Turbines”, IEEE Transaction on Energy Conversion, Vol 22, N°2, June 2007 [6] M. A. Tankari, B. Dakyo, C. Nichita, “Improved sizing method of storage units for hybrid wind-diesel powered system”, Power Electronics and Motion Control Conference, 2008. EPE-PEMC 2008. 13th, 1-3 Sept. 2008 Page(s) : 1911 – 1917 [7] J. Max, “Pratique du filtrage, filtrage analogique”, Techniques de l’ingénieur, R1102 [8] G. Lissorgues, “Filtres actifs Synthèse et réalisation”, Techniques de l’ingénieur, E 115 [9] H.Bindner and al., “Lifetime Modelling of Lead Acid Batteries”, Risø National Laboratory, Roskilde Denmark, April 2005. [10] A.D. Hansen and al.; “Models for a standalone PV system”; Riso National Laboratory, Roskilde December 2000 [11] T. Christen, Martin W. Carlen, “Theory of Ragone plots”, Journal of Power Sources 91, pp210–216, March 2000 [12] S. H. Baek and al., “Fatigue life prediction based on the rainflow cycle counting method for the end beam of a freight car bogie”, International Journal of Automotive Technology, Vol. 9, No. 1, pp. 95-101 (2008) [13] Http://www.ceere.org/rerl/
