1

Energy Eciency Comparison of Electrical Systems for Wind Turbines

Andreas Petersson and Stefan Lundberg
Department of Electric Power Engineering Chalmers University of Technology SE-412 96 G oteborg, Sweden e-mail: andreas.petersson, stefan.lundberg@elteknik.chalmers.se

Power [MW]

Abstract The eciency of the electrical system of doublyfed induction generators, for variable speed wind turbines, in comparison to other electrical systems has been investigated. It has been found that there is a possibility to gain a few percent in energy eciency using a doubly-fed induction generator compared to an ordinary induction generator equipped with an full-power inverter.

a) 1

0.5

I. Introduction IND TURBINES (WT) can operate with either xed speed or variable speed. For xed-speed WT, the generator (induction generator) is directly connected to the grid. Since the speed is almost xed to the grid frequency, and most certainly not controllable, it is not possible to store the turbulence of the wind in form of rotational energy. Therefore, for a xed-speed system the turbulence of the wind will result in power variations, and thus aect the power quality of the grid [1]. For a variable-speed WT the generator is controlled by power electronics equipment, which makes it possible to control the rotor speed. By this way the power uctuations caused by wind variations can be absorbed with rotor speed variations [2]. Hence, the power quality impact caused by variable-speed wind turbine can be improved compared to the xed-speed turbine [3]. In [4] Grauers has thoroughly described and analyzed the losses of the permanent magnet synchronous generators (PMSG) for a WT application. When the stator of the IG is connected to an inverter (or the grid) the system is referred to as singly-fed induction generator (SFIG). Another possibility for the IG is to have the stator circuit directly connected to the grid and the rotor circuit connected to an inverter, which is referred to as doubly-fed induction generator (DFIG). The purpose of this paper is to make a theoretical investigation of the energy eciency for the electrical systems of WT with DFIG compared to a WT with xed-speed IG (FSIG), SFIG or PMSM. The rotor-speed range and the stator-to-rotor turns ratio of the IG are important aspects that will be studied. II. Wind Energy Conversion Theory Some of the available power in the wind is converted by the rotor blades to mechanical power acting on the rotor shaft of the WT. For steady-state calculations of the mechanical power from a wind turbine the so called Cp ()curve can be used. The mechanical power, Pmech , can be

0 b) Rotor Speed [rpm]

10

15 Wind Speed [m/s]

20

25

25 20 15 10 5 10 15 Wind Speed [m/s] 20 25

Fig. 1. Typical characteristic for a variable speed wind turbine. a) Mechanical power as a function of wind speed. b) Rotor speed as a function of wind speed.

determined by [5] Pmech = 1 Ar Cp (, )w3 2 r rr w (1) (2)

where Cp is the power coecient, is the pitch angle, is the tip speed ratio, w is the wind speed, r is the rotor speed (on the low speed side of the gear box), rr is the rotor plane radius, is the air density and Ar is the area swept by the rotor. In this paper it is assumed that the WT is equipped with pitch control that limit the input power. Fig. 1 shows an example of how the mechanical power, derived from the Cp ()-curve, and the rotor speed vary with the wind speed for a variable speed WT. The rotor speed in the variable speed range is controlled in order to keep the optimal tip speed ratio, . The rotor-speed range is an important design parameter for a WT equipped with a DFIG. The two most commonly used probability density functions to describe the wind speed are the Weibull and the Rayleigh functions [5]. The Weibull distribution is de-

replacements Grid Inverter IG/SG r Inverter IM r

PSfrag replacements
Fig. 2. SFIM.

Grid scribed by the following probability density function k w k1 (w/c)k e (3) c c where k is a shape parameter, c is a scale parameter and w is the wind speed. Thus, the mean wind speed (or the expected wind speed during a year) can be calculated from f (w) =

Fig. 3. DFIM.

wmean =
0

wf (w)dw =

1 c k k

(4)

where is Eulers gamma function, i.e.,

(z ) =
0

tz1 et dt.

(5)

If the shape parameter equals 2, the Weibull distribution is known as the Rayleigh distribution. The advantage of the Rayleigh distribution is that it only depends on the mean wind speed. For the Rayleigh distribution the scale factor, c, given the mean wind speed can be found from (k =2, and ) ( 1 ) = 2 2 c = wmean . (6) A wind speed of 5.4 m/s and 7.2 m/s correspond, respectively to a medium and a high wind site in Sweden respectively, according to [6]. III. Electrical System Models for Variable-Speed Wind Turbines A. General Description In this section two variable speed systems will be presented: a stator-fed generator and a doubly-fed IG. The rst system, presented in Fig. 2, consists of a WT equipped with an inverter connected to the stator of the generator. The generator could either be a SFIG or a PMSM. Since the PMSG can be designed with multiple poles there is no need for a gear-box in the PMSG-system, but in the case of an IG the gear-box is designed so that maximum rotor speed corresponds to rated speed of the IG. For the SFIG and the PMSG a major advantage is its welldeveloped and robust control [7], [8], [9]. The xed-speed WT with an IG (FSIG) can be seen as a special case of the variable-speed SFIG-system with no inverter, i.e., the IG is directly connected to the grid. The FSIG-system is equipped with two generators one that can handle the full power and one smaller generator with lower synchronous speed (low wind speed operation). The third system, see Fig. 3, consists of a wind turbine with a variable-speed constant-frequency induction generator (DFIG). This means that the stator is directly connected to the grid while the rotor winding is connected via

slip rings to an inverter. The inverter is designed so that the IG can operate in a limited variable-speed range. The speed range, i.e., the slip, is approximately determined by the ratio between the stator and rotor voltage. The gearbox ratio is set so that the nominal speed of IG corresponds to the middle value of the rotor-speed range of the WT. This is done on order to minimize the size of the inverter, which will vary with the rotor-speed range. With this inverter it is possible to control the speed (or the torque) and also the reactive power on the stator side of the IG. The transformer between the inverter and rotor circuit in Fig. 3 is to indicate and highlight the stator-to-rotor turns ratio. The stator-to-rotor turns ratio can be designed so that maximum voltage of the inverter corresponds to the desired maximum rotor voltage to get the desired slip. This implies that it is possible to reduce the conducting losses of the inverter. The transformer can be treated as ideal since it is actually included in the IG. (Note that there is a possibility to use a higher stator voltage, lets say 2.8 kV but have a maximum rotor voltage of 690 V and obtain the same eect as with 690 V stator voltage and a turns ratio of 1:4.) However, here only the case with a 690 V nominal stator voltage has been considered. The control of the DFIG is more complicated and has all the limitations that the line-fed synchronous generator has, e.g., starting problem, synchronization and oscillatory transients [7]. In [10] it is shown that the DFIG dynamics contains poorly damped eigenvalues with the natural frequency near the line frequency. This causes oscillations in the air-gap ux. There exist a variant of the DFIG method which uses external rotor resistances which can be controlled (compare to slip power recovery). Some of the drawbacks of this method are that it is not possible to decrease the rotor speed below synchronous speed, energy is unnecessary dissipated in the external rotor resistances and that it is not possible to control reactive power. B. The Induction Generator A commonly used model for the IG is the Park model, [11], (or the equivalent T-model). Using space-vectors the Park model, with the inclusion of iron losses, can be expressed in synchronous coordinates as vs vr = Rs is + = Rr ir + d s dt d r dt + j1 s + j(1 r ) r (7) (8)

Rs + vs s -

Ls

is s

is r

s replacements Lr PSfrag Rr j r r + + -

T1

T2

T3

VCE 0 rCE

Lm

Rm is m,R

vs r T4 T5 T6

VT 0 rT

Fig. 4. Park model of the IG in stator coordinates (1 = 0).

(9) + j 1 m dt where the following notation is used. vs stator voltage m air-gap ux vr rotor voltage Rs stator resistance stator current Rr rotor resistance is rotor current Rm magnitizing resistance ir magnitizingsynchronous im,R 1 resistance current angular speed s stator ux r rotor angular speed rotor ux r An equivalent circuit of the IG in stator coordinates can be seen in Fig. 4. Since the model is described in synchronous coordinates, the variables will be dc quantities in the steady state and the derivatives will be zero. The stator, rotor and air-gap uxes in (7) to (9) are given as Rm im,R + s r m = = = Ls is + Lm ir + Lm im,R = Ls is + m (10) Lm is + Lr ir + Lm im,R = Lr ir + m (11) Lm is + Lm ir + Lm im,R (12)

d m

Fig. 5. Inverter scheme.

where Lm is the magnitizing inductance, Ls is the stator inductance, Lr is the rotor inductance, Ls is the stator leakage inductance and Lr is the rotor leakage inductance. The resistive losses of the IG is Ploss = 3(Rs |is |2 + Rr |ir |2 + Rm |im,R |2 ) (13)

and it is possible to express the electro mechanical torque, Te , as (14) Te = 3np Im m i r = 3np Im r ir where np is the number of pole pairs. If the IG is equipped with slip rings, the voltage drop across the slip rings has been neglected. Variations in the magnetizing resistance, Rm , due to applied stator voltage and frequency have also been neglected. For the IG:s, used in this paper, operated at 690 V 50 Hz the following parameters are used: 1 MW: Rs = 3.1 m, Rr = 4.6 m, Rm = 85 , Ls = 0.15 mH, Lr = 0.15 mH, Lm = 7 mH and np = 2 0.4 MW: Rs = 4.0 m, Rr = 1.0 m, Rm = 200 , Ls = 0.41 mH, Lr = 0.14 mH, Lm = 12 mH and np = 3. C. Inverter In order to be able to feed the IG with a variable voltage and frequency source, the IG can be connected to a pulsewidth modulated (PWM) inverter. In Fig. 5, an equivalent

circuit of the inverter is drawn, where each transistor, T1 to T6, is equipped with a reverse diode. A PWM circuit switches on and o the transistors. The duty cycle of the transistor and the diode determines whether the transistor or a diode is conducting in a transistor leg (e.g., T1 and T4). The losses of the inverter can be divided into switching losses and conducting losses. The switching losses of the transistors are the turn-on and turn-o losses. For the diode the switching losses mainly consist of turn-o losses [12], i.e. reverse-recovery energy. The turn-on and turn-o losses for the transistor and the reverse-recovery energy loss for a diode can be found from data sheets. The conducting losses arise from the current through the transistors and diodes. The transistor and the diode can be modeled as constant voltage drops, VCE 0 and VT 0 , and a resistance in series, rCE and rT , see Fig. 5. Simplied expressions of the transistors and diodes conducting losses, for a transistor leg, are (with a third harmonic voltage injection) [13] 2 VCE 0 Irms 2 Irms VCE 0 mi cos() rCE Irms + + Pc,T = 2 6 2 2 rCE Irms mi 4rCE Irms mi cos() + (15) 3 cos()6 45 3 2 VT 0 Irms 2 Irms VT 0 mi cos() rT Irms Pc,D = + 2 6 2 2 r T I mi 4rT Irms mi cos() rms + (16) 3 cos()6 45 3 where Irms is the root mean square (RMS) value of the (sinusoidal) current to the grid or the generator, mi is the modulation index, and is the phase shift between the voltage and the current. Since, for the values in this paper [14], [15], [16] (see Table I for actual values), r = rCE rT and V = VCEO VT O it is possible to model the transistor and the diode with the same model. The conduction losses can, with the above mentioned approximation, be written as 2V Irms 2 2 Pc = Pc,T + Pc,D = + rIrms . (17) A reasonable assumption is that the switching losses of the transistor is proportional to the current [13]. This implies

that the switching losses from the transistor and the inverse diode can be expressed as 2 2Irms fsw (18) Ps,T = (Eon + Eo ) IC,nom Ps,D = Err fsw (19) where Eon and Eo is the turn-on and turn-o energy losses respectively, for the transistor, Err is the reverse recovery energy for the diode and IC,nom is the nominal current through the transistor. The total losses from the three transistor legs of the inverter becomes Ploss = 3(Pc + Ps,T + Ps,D ). (20)

IG Losses [%]

2 1 0

10

15 Wind Speed [m/s]

20

25

Fig. 6. Induction generator losses, in percent of maximum shaft power, at dierent wind speeds. FSIG (solid), SFIG (dashed) and DFIG (dash-dotted).

The back-to-back inverter can be seen as two inverters which are connected together: the machine-side inverter, (MSI), and the grid-side inverter (GSI). For the MSI the current through the valves, Irms , are the stator current for the SFIG system or the rotor current for the DFIG system. One way of calculating Irms for the GSI is by using the current that produce the active power in the machine, adjusted with the ratio between machine-side voltage and the grid voltage. The reactive current is assumed to be stored in the dc link capacitor. Thus it is now possible to calculate the losses of the back-to-back inverter as Ploss,inverter = Ploss,GSI + Ploss,MSI . (21)

0 Power [%] 10 20 30 40 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm]
Fig. 7. Maximum active power (solid) and maximum reactive power (dashed) that the inverter supply the DFIG. Active power is in percent of maximum active power and reactive power is in percent of maximum reactive power respectively that is handled by the inverter in the SFIG system.

In this paper the switching frequency is set to 5 kHz. D. Gear-Box Losses One way to estimate the gear-box losses, Ploss,GB , is, [17], Ploss,GB = Plowspeed + Pn r rn (22)

where is the gear mesh losses constant and is a friction constant. According to [4], for a 1-MW gear box, the constants = 0.02 and = 0.005 are reasonable. E. Total System Losses When calculating the losses for the system we will take into account the IG losses, gear-box losses, as well as machine-side and grid-side inverter losses. The total system losses become Ploss = Ploss,GB + Ploss,IG + Ploss,GSI + Ploss,MSI (23)

speed. As can be seen in the gure the losses of the DFIM are higher than those of the SFIM for low wind speeds. The reason for this is that the ux level of the SFIM-system can be optimized while for the DFIM the ux level is xed to the stator voltage. Fig. 7 shows the maximum power and the maximum reactive power that is fed to the DFIG by the inverter as a function of the rotor speed range. Since the stator current, stator voltage, rotor current and rotor voltage for dierent wind speeds have been calculated, it is possible to calculate the losses of the backto-back inverter according to Section III-C. The transistor that is chosen, from Table I, depends on the maximum peak current through the valves. In Fig. 8 the inverters total losses as a function of the wind speed can be seen. It is possible to reduce the current through the inverter by the stator-to-rotor turns ratio, indicated by the transformer in Fig. 3. This implies that it is possible to reduce the conducting losses of the inverter. Therefore the inverter losses for the DFIG are calculated and shown for three dierent turns ratio in the gure. However, one thing that has to be kept in mind is that the variable-speed range becomes very limited for the system with a stator-to-rotor turns ratio of 1:11, with low inverter losses. The total losses of the system, as a function of the wind speed, is shown in Fig. 9. The mean value of the produced power, during a year,

where Ploss,IG are the losses of the IG. The losses of the slip-rings for the DFIG-system and friction losses of the IG are neglected. IV. Result of Calculations Steady-state calculations will be carried out in this section in order to determine the losses of the two systems. For the DFIG system the reactive power at the stator has been set to 0 VAr. In Fig. 6 the losses of the generators for the FSIG, the SFIG and the DFIG cases, are plotted as a function of wind

Inverter Losses [%]

SFIG

2 1 0 5 10 15 Wind Speed [m/s] 20

DFIG 1:1 DFIG 1:3 DFIG 1:11

25

Energy Gain [%] Energy Gain [%]

Fig. 8. Inverter losses at dierent wind speeds in percent of maximum shaft power.

10 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm] b) 5

Energy Gain [%]

(24)

a)

Total Losses [%]

8 6 4 2 0 5 10 15 20 Wind Speed [m/s]

SFIG DFIG 1:1 DFIG 1:3 FSIG

5 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm] c) 2 0 2 4 6 8 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm]
Fig. 10. Gain in energy for a DFIG-system, for mean wind speeds of 5.4 m/s (solid), 6.8 m/s (dashed) and 7.2 m/s (dotted), as a function of the rotor speed range. The gain in energy is in comparisons with a) FSIG. b) SFIG. c) PMSG.

25

Fig. 9. Total losses of the system as a function of wind speed.

can be found from

Pmean =
0

P (w)f (w)dw

where f (w) is the probability density function. The produced energy can be found by multiplication of time. In Fig. 10 the gain in energy for a DFIG-system compared to FSIG, SFIG, and PMSG, for dierent mean wind speeds (5.4 m/s, 6.8 m/s and 7.2 m/s), as a function of the rotor speed range. The stator-to-rotor turns ratio, for the DFIG, is adjusted so that maximum rotor voltage is 75 % of the rated voltage, i.e., 75 % of 690 V. The average eciency, with a mean wind speed of 6.8 m/s, for the PMSG generator is taken from [4]. The eciency is scaled linearly with the mean wind speed. The inverter losses of the PMSGsystem are assumed similar to the SFIG-system. It can be seen in Fig. 10a that the FSIG-systems performs better than the DFIG-system. This is due to that the FSIG-system is equipped with two generators that can operate at two dierent synchronous speeds. For the comparison to the SFIG-system it is possible to gain a few percent if the variable-speed range is set properly. In comparison to the PMSG-system there is no gain in energy (for medium mean wind speeds), i.e., the systems produce the same amount of energy. For high-mean wind-speed sites there might be a slight gain in energy. It is possible to lower the magnetizing losses of the DFIGsystem and increase the gain in energy even further. This can be done by: 1) By short circuit the stator of the IG, at low wind speeds, and take all power out through the inverter. 2) By having the stator in -connection at high wind speeds and in Y-connection at low wind speeds. For

the rst case it is possible to gain about 0.3 % extra energy, for a high wind speed site, and up to 0.9 % extra energy, for a low wind speed site. In Fig. 11 the increased gain in energy by reducing the magnetizing losses, by the two above mention methods, are presented. The gain in energy for the system with a Y--coupling performs approximately 30 % better than the system with short circuited stator. The break-even point of the total losses or the rated values of the inverter determines the switch over point, e.g. the Y- coupling. It can also be seen in Fig. 11 that the Y--coupling has approximately the same gain in energy independent of the rotor speed range. One reason for this is due to the fact that the inverter at low wind speeds only handles a fraction of the total power compared to when the stator is short circuited and the inverter handles the total power. V. Conclusions In this paper the gain in total energy produced, during a year, of a doubly-fed induction generator system compared to a singly-fed generator system, for a WT application, has been studied. It was found that if the range of the variable speed is set properly, there is the possibility to gain a

Energy Gain [%]

a) 1.5 1 0.5

[6]

[7] [8] [9] [10]

Energy Gain [%]

0 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm] b) 1.5 1 0.5 0 2425 2225 2025 1825 1625 1425 1225 Rotor Speed Range [rpm]

[11] [12] [13] [14] [15] [16] [17]

Fig. 11. Increased gain in energy, for mean wind speeds of 5.4 m/s (solid), 6.8 m/s (dashed) and 7.2 m/s (dotted), as a function of the rotor speed range, for a DFIG-system when it is equipped with: a) Short circuited stator at low wind speeds. b) Y- coupling.

T. Thiringer and J. Linders, Control by variable rotor speed of a xed-pitch wind turbine operating in a wide speed range, IEEE Trans. on Energy Conversion, vol. 8, no. 3, pp. 5206, Sept. 1993. W. Leonhard, Control of Electrical Drives, Springer-Verlag, 2nd edition, 1996. I. Boldea and S. A. Nasar, Electric Drives, CRC Press LCC, 1999. L. Harnefors and H.-P. Nee, Control of Variable-Speed Drives, Electrical Machines and Power Electronics, Dept. of Electrical Engineering, Royal Institute of Technology, Stockholm, 2000. M. Heller and W. Schumacher, Stability analysis of doublyfed induction machines in stator ux reference frame, in Proceedings of 7th European Conference on Power Electronics and Applications, Brussels, Belgium, 1997, pp. 707710 vol. 2, EPE Assoc. P. K. Kov acs, Transient Phenomena in Electrical Machines, Elsevier, 1984. K. Thorborg, Power Electronics in Theory and Practice, Studentlitteratur, 1993. F. Abrahamsen, Energy Optimal Control of Induction Motor Drives, Ph.D thesis, Aalborg University, Institute of Energy Technology, Aalborg, Denmark, February 2000. Semikron, SKiiP 2403GB122-4DL, Data sheet, March 2001. Semikron, SKiiP 1203GB122-2DW, Data sheet, March 2001. Semikron, SKiiP 603GD122-3DUW, Data sheet, March 2001. A. Grauers, Synchronous Generator and Frequency Converter in Wind Turbine Applications: System Design and Eciency, Licentiate thesis, Chalmers University of Technology, Department of Electric Power Enginering, G oteborg, Sweden, May 1994.

Appendix
TABLE I Inverter Data. Inverter Characteristics 1 (IGBT and inverse diode) Nominal current IC,nom 1200 A Operating dc-link voltage VCC 1000 V VCEO 1.2 V Lead resistance (IGBT) rCE 0.8 m Turn-on and turn-o Eon + Eo 733 mJ energy (IGBT) VT O 1.2 V Lead resistance (diode) rT 0.7 m Reverse recovery Err 163 mJ energy (diode) Inverter Characteristics 2 (IGBT and inverse diode) Nominal current IC,nom 600 A Operating dc-link voltage VCC 1000 V VCEO 1.2 V Lead resistance (IGBT) rCE 1.6 m Turn-on and turn-o Eon + Eo 367 mJ energy (IGBT) VT O 1.2 V Lead resistance (diode) rT 1.3 m Reverse recovery Err 81.3 mJ energy (diode) Inverter Characteristics 3 (IGBT and inverse diode) Nominal current IC,nom 300 A Operating dc-link voltage VCC 1000 V VCEO 1.2 V Lead resistance (IGBT) rCE 3.1 m Turn-on and turn-o Eon + Eo 183 mJ energy (IGBT) VT O 1.2 V Lead resistance (diode) rT 2.7 m Reverse recovery Err 40 mJ energy (diode)

few percents in energy eciency compared to a single-fed induction generator. However a direct-driven permanentmagnet synchronous generator will approximately perform as the doubly-fed induction generator system. For a site with a high mean wind, the doubly-fed induction generator system might perform slightly better than the permanentmagnet synchronous generator. If the stator of the doublyfed induction generator is equipped with a connection device for Y--connections the gain in energy can be increased even further. The stator-to-rotor turns ratio is an important design parameter for lowering the losses of the doubly-fed induction generator system. Acknowledgments The nancial support provided by the Swedish National Energy Agency and ABB Power Systems is gratefully acknowledged. References
[1] M.P. Papadopoulos, S.A. Papathanassiou, N.G. Boulaxis, and S.T. Tentzerakis, Voltage quality change by grid-connected wind turbines, in European Wind Energy Conference, Nice, France, 1999, pp. 783785. T. Petru and T. Thiringer, Active icker reduction from a seabased 2.5 MW wind park connected to a weak grid, in 2000 IEEE Nordic Workshop on Power and Industrial Electronics (NORpie/2000), Aalborg, Denmark, 2000, pp. 711. A. Larsson, P. Srensen, and Santjer F., Grid impact of variable speed wind turbines, in European Wind Energy Conference and Exhibition (EWEC 99), Nice, France, 1999. A. Grauers, Design of Direct-Driven Permanent-Magnet Generators for Wind Turbines, Ph.D thesis, Chalmers University of Technology, Department of Electric Power Enginering, G oteborg, Sweden, October 1996. G.L. Johnson, Wind Energy Systems, Prentice-Hall, 1985.

[2]

[3] [4]

[5]