Controller Design for an Induction Generator Driven
by a Variable-Speed Wind Turbine
Woei-Luen Chen, Member, IEEE, and Yuan-Yih Hsu, Senior Member, IEEE
Abstract—This paper presents the modeling, controller design and a steady-state analysis algorithm for a wind-driven induction generator system. An output feedback linear quadratic controller is designed for the static synchronous compensator (STATCOM) and the variable blade pitch in a wind energy conversion system (WECS) in order to reach the voltage and mechanical power con-trol under both grid-connection and islanding conditions. A two-reference-frame model is proposed to decouple the STATCOM real and reactive power control loops for the output feedback con-troller. To ensure zero steady-state voltage errors for the output feedback controller, the integrals of load bus voltage deviation and dc-capacitor voltage deviation are employed as the additional state variables. Pole-placement technique is used to determine a proper weighting matrix for the linear quadratic controller such that sat-isfactory damping characteristics can be achieved for the closed-loop system. Effects of various system disturbances on the dynamic performance have been simulated, and the results reveal that the proposed controller is effective in regulating the load voltage and stabilizing the generator rotating speed for the WECS either con-nected with or disconcon-nected from the power grid. In addition, proper steady-state operating points for an isolated induction gen-erator can be determined by the proposed steady-state analysis algorithm. Constant output frequency control using the derived steady-state characteristics of the isolated induction generator is then demonstrated in this paper.
Index Terms—Induction generator (IG), static synchronous compensator (STATCOM), voltage regulation, wind energy con-version system (WECS), wind turbine.
NOMENCLATURE General
e, E Inverter output voltage in instantaneous and vector notation.
fo Output frequency of the induction generator (IG)
terminal voltage.
I Identity matrix.
i, I Current in instantaneous and vector notation.
Ks, Ko State and output feedback gain matrices.
Pm, Pg Input mechanical power and output real power for
the IG.
r Resistance.
Te, Tm Electrical and mechanical torques.
u Input vector.
Manuscript received May 25, 2005; revised January 14, 2006. This work was supported by the National Science Council Taiwan, R.O.C., China under Contract NSC-93-2213-E-002-011. Paper no. TEC-00173-2005.
W.-L. Chen is with the Department of Electrical Engineering, National Taiwan University, and also with the Department of Electronic Engineer-ing, Hwa Hsia Institute of Technology, Taipei 235, Taiwan, R.O.C. (e-mail: [email protected]).
Y.-Y. Hsu is with the Department of Electrical Engineering, National Taiwan University, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected]).
Digital Object Identifier 10.1109/TEC.2006.875478
v, V, V Voltage in instantaneous, magnitude, and vector notation.
Vw Wind speed.
X Reactance.
x State vector.
y Output vector.
β Pitch angle of the wind turbine.
θ Relative angular displacement for a specific vector, where−π<θ= π.<
φ Flux linkage.
ω Angular frequency (angular speed).
0 Zero matrix.
Subscripts
a Vector or matrix for the augmented system. b Base quantity.
d, q Quantities in d-axis and q-axis. dc dc link.
f Coupling transformer and filter for the STATCOM. FC Fixed shunt capacitor.
L Load.
s, r Stator and rotor.
T Wind turbine.
TL Transmission line. ∞ Grid (infinite bus). 0 Initial operating point. Superscripts
T Matrix transpose operator. ∗ Reference value.
I. INTRODUCTION
I
NDUCTION generators are being increasingly utilized in a wind energy conversion system (WECS) since they are rel-atively inexpensive, rigid, and require low maintenance. How-ever, the impact of ever-changing wind speed on power quality, coupled with the need of excitation current for induction gen-erator (IG), make the mechanical power control and voltage regulation indispensable to the wind-driven induction generator system. By far, the most effective way of controlling the me-chanical power captured by the wind turbine is to adjust the rotor blade pitch angle. Blade pitch is analogous to the throttle valve in conventional steam turbines, except that the speed of control is much faster than the governor control in a steam turbine [1], [2]. It can be employed to regulate mechanical power input and real power output of the WECS. However, the reactive power required by the IG can be provided by a shunt capacitor bank, but it may cause excessive overvoltage during disconnection. More-over, the amount of capacitance required for excitation varies 0885-8969/$20.00 © 2006 IEEEwith the generator speed [3]. Thus, if a fixed shunt capacitor is connected across the terminals of the IG, the terminal voltage will vary with generator speed. To achieve continuous voltage regulation under varying system conditions, static synchronous compensators (STATCOMs), have been employed in the litera-tures [4]–[8]. The basic principle of a STATCOM installed in a power system is to generate a controllable ac voltage behind a coupling transformer and a filter by a voltage-sourced inverter (VSI) connected to a dc capacitor. The output voltage of the VSI can be controlled to be greater than the line voltage in order to provide reactive power to the wind-driven IG.
An application of the linearized multivariable output feed-back technique to the design of a voltage and mechanical power regulator for a wind turbine-induction generator around a spe-cific operating point was given in [1], [9]. Nevertheless, for the WECS, it is impractical and uneconomical to fix the generator output power (or rotor speed) at a specific operating point since the wind speed changes all the time. Besides, the controller de-sign based on the constant output power (or rotor speed) consid-erations under grid-connection condition would fail in islanding operation.
In this paper, an output feedback controller is designed to regulate system voltage and mechanical power under both grid-connection and islanding conditions. The major features of the proposed output feedback controller are summarized as follows. 1) A systematic approach based on pole placement technique has been applied to determine proper weighting matrix for the linear quadratic controller such that the eigenvalues for the closed-loop system can be shifted to desirable loca-tions. As a result, satisfactory transient excursion suppres-sion in system voltages and generator speed can be reached by the controller when the WECS is subjected to major disturbances.
2) The integrals of load bus voltage deviation and dc-capacitor voltage deviation are employed as additional state variables to eliminate the steady-state voltage errors for the output feedback controller.
3) A control scheme with two reference frames is proposed to reach the decoupled control for the STATCOM real and reactive power control loops in case of voltage fluc-tuations, no matter what kind of disturbances may take place.
4) A simple algorithm is developed to determine proper steady-state operating points for an isolated IG. Using the relationship between the generator speed and the output power at various frequencies, a proportional-integral con-troller is proposed to achieve constant frequency control for an isolated IG with sufficient mechanical power.
II. SYSTEMMODELS
Fig. 1 depicts the one-line diagram of an induction generator driven by a variable-speed wind turbine and connected to a grid through a transmission line. An output feedback controller is utilized to control the wind turbine mechanical power and the system reactive power through the variable blade pitch and the STATCOM, respectively. The reactive power required by the IG
Fig. 1. System configuration.
in steady-state operating condition is supplied by a fixed shunt capacitor bank, as shown in Fig. 1. To maintain constant load bus voltage (VL) under disturbance conditions, a STATCOM,
which is capable of adjusting its output voltage and reactive power output based on system requirements, is employed. The STATCOM is connected to the load bus through a coupling transformer and a filter. Note that a constant impedance load is assumed in Fig. 1.
A. Induction Generator Model
The per unit flux-linkages for the stator and rotor circuits of the induction generator described in d- and q-axes are as follows [9], [10]: ˙ φds = ωb(vdL + rsids) + ωsφq s (1) ˙ φq s= ωb(vq L+ rsiq s)− ωsφds (2) ˙ φdr = ωb(vdr− rridr) + (ωs− ωr)φq r (3) ˙ φq r= ωb(vq r− rriq r)− (ωs− ωr)φdr (4)
where a synchronous reference frame rotating at the electrical angular speed corresponding to the fundamental frequency of the grid voltage, herein denoted as ωs, is adopted.
The electromechanical torque in per unit can be written in terms of stator flux linkages and currents as
Te = φdsiq s− φq sids. (5)
The per unit rotor acceleration is given by ˙
ωru = 1 2HT
(Tm− Te− DTωur) (6)
where Tm is the per unit mechanical torque of the wind
tur-bine, and HT and DT are the equivalent inertia constant and
the equivalent damping constant of the wind turbine-induction generator system, respectively.
B. Wind Turbine Model
The mechanical power output of a wind turbine can be written as [2] Pm = 1 2ρACpV 3 w (7)
where ρ is the air mass density, Vw is the wind speed, A is
the rotor swept area, and Cp is a power coefficient representing
the fraction of power extracted from the aerodynamic power in the wind by a practical wind turbine.
The power coefficient Cp varies with the wind speed, the
rotational speed of the turbine, and the turbine blade parameters. The MOD-2 wind turbine model [11], [12] with the following closed-form approximate relationship for Cp is used:
Cp = 1 2 R γ − 0.022β 2− 5.6 e−0. 17Rγ . (8)
The tip speed ratio γ is defined as γ = ωTR Vw
(9) where ωT is the rotating speed of the wind turbine as shown in
Fig. 1.
It is observed from (7)–(9) that the mechanical power output of a wind turbine is related to the turbine speed ωT, wind speed
Vw ,and the pitch angle β. An increase in the pitch angle β, which
results in a decrease in the power coefficient Cp, is efficient to
control the mechanical power of the wind turbine when Pm
continues to increase with increasing wind speed.
In this work, the initial pitch angle (β0) is chosen to be 13.46◦
such that the wind turbine delivers a mechanical power of 0.81 per unit (p.u.) for a 30 miles/h (mph) wind at hub height. A hydraulic pitch angle actuator with a rate limit of±10◦/s and a transport delay of 50 ms [11], [12] is used in this paper. C. STATCOM Model
For a balanced three-phase system, the STATCOM model can be described in per unit using the variables in d- and q-axes synchronous reference frame as [4]
˙ide=− ωbrf Xf ide+ ωsiq e+ ωb Xf (vdL − ed) (10) ˙iq e=− ωbrf Xf iq e− ωside+ ωb Xf (vq L− eq). (11)
In Fig. 1, the instantaneous powers at the ac and dc sides of the voltage-sourced inverter are equal, giving the following power balance equation:
vdcidc= edide+ eqiq e. (12)
The per unit dc-side circuit equation is ˙vdc= 1 Cdc idc− vdc rdc (13) where rdcis used to represent the inverter switching loss.
The per unit equations of the fixed shunt capacitor, transmis-sion line, and local load models are given in the Appendix.
Fig. 2. Power flow between STATCOM and the load bus.
III. DERIVATION OF THEAUGMENTEDSTATEEQUATIONS In order todetermine proper control signal for the STATCOM through a systematic design approach, the dynamic system equa-tions as given in Section II are linearized around a nominal operating point in the form as
˙
∆x = A· ∆x + B · ∆u
∆y = C· ∆x (14a)
where
∆x = [∆vdL ∆vq L ∆idL ∆iq L ∆ωur ∆vdc ∆φq r
∆φdr ∆φds ∆φq s ∆idTL ∆iq TL ∆ide ∆iq e]T
(14b) is the state vector. It remains now to define the control action ∆u and the output vector ∆y based on the two-reference-frame transformation described below.
A. Decoupled Control Using the Two-Reference-Frame Transformation
The purpose of the STATCOM controller is to make use of the real power and reactive power exchange between the STATCOM and the ac system to achieve good regulation of dc-capacitor voltage and the load bus voltage. In other words, STATCOM control is closely linked to the power flow at the STATCOM terminal, as shown in Fig. 2. It is well known [4], [5] that the instantaneous real and reactive power, through a coupling path to the STATCOM, at the load bus can be represented as follows: Pe = vdLide+ vq Liq e (15a)
Qe = vq Lide− vdLiq e. (15b)
Note that the d–q quantities of the STATCOM current (ide
and iq e), STATCOM voltage (ed and eq), and load bus voltage
(vdL and vq L) in (10), (11), and (15) are derived by choosing
the vector for grid bus voltage V∞as the d-axis. To decouple the STATCOM real and reactive power control loops, a new reference frame which uses the vector for load bus voltage VL
as the new d-axis (denoted as d-axis) is defined in Fig. 3 [4]. Note that the angles θe, θL, and θi in Fig. 3 are the phase
angles for the voltages E, VL, and STATCOM current Ie, with
respect to the original d-axis, respectively. On the other hand, the angles θeand θiin Fig. 3 are the phase angles for the voltage
E and STATCOM current Ie with respect to the new d-axis,
respectively. As shown in Fig. 4, the voltages ed and eq and currents ideand iq ein the new reference frame are related to the
Fig. 3. Transformation between the two reference frames illustrated by trigonometric relationships.
Fig. 4. Block diagram of the proposed output feedback controller.
voltages edand eqand currents idand iqin the original reference
frame through the transformation matrix T as described below. ide iq e = T ide iq e and ed eq = T ed eq (16) where T = cos θL sin θL − sin θL cos θL .
Using the d–q quantities vdL (which is equal to VL), ide, and
iq e in the new reference frame, the real and reactive powers in (15) can be written as
Pe = vdL ide (17a)
Qe =−vdLiq e. (17b)
As (17) shows, the d-axis current component ide accounts for the instantaneous real power and the q-axis current compo-nent iq eis the instantaneous reactive current. Thus, STATCOM control design is simplified to a great extent with this new ref-erence frame since the load bus voltage regulation is achievable through the control of the reactive power Qewhich only related
to the q-axis current iq e.
Evidently, the transformation matrix T as described in (16) and its inverse are needed to transform certain system variables from the synchronous reference frame to the other in order to decouple STATCOM real and reactive power control loops for the proposed controller, as shown in Fig. 4. Thus, the control
inputs for the state equations in (14a) are
∆u = [∆ed ∆eq ∆β]T (18)
and the output vector is as follows:
∆y =∆VL ∆vdc ∆ωru ∆ide ∆iq e
T
= C∆x (19)
where the output matrix C is given by
C =CT1 CT2 CT3 CT4 CT5T
and the submatrices C1, C2, C3, C4, and C5 are defined
as ∆VL = C1∆x, ∆vdc= C2∆x, ∆ωur = C3∆x, ∆ide =
C4∆x, and ∆iq e = C5∆x, respectively.
Note that the mechanical power can be conditioned by a variable blade pitch angle β in (18) through a controller with IG speed error ∆ωu
r as the primary stabilizing signal in the output
vector in (19).
B. Elimination of Steady-State Errors Through the Augmented State Equations
To ensure zero steady-state errors for load bus voltage VL
and dc-capacitor voltage vdc, two integral terms
∆VLdt and
∆vdcdt are required in addition to the state variables in (14b).
Furthermore, the operating point of the IG changes with wind speed and load and different reference values for ∆iq e and ∆ide may be required by the STATCOM to achieve good reg-ulation of load bus voltage VL and dc-capacitor voltage vdc.
In order for the STATCOM currents ∆iq e and ∆ide to fol-low the reference values ∆i∗q e and ∆i∗de and ensure tight reg-ulations for δiq e= ∆i∗q e− ∆iq e and δide = ∆i∗de− ∆ide, we need two additional state variables δiq edt and δidedt. In the present work, the integral terms ∆VLdt (which is
re-lated toδiq edt, since
∆VLdt is proportional to ∆i∗q e [4])
and ∆vdcdt (which is related to
δidedt, since ∆vdcdt
is proportional to ∆i∗de[4]) are employed as the additional state variables. With the four integral terms ∆VLdt,
∆vdcdt,
∆VLdt, and
∆vdcdt as the additional state variables,
we have the augmented state equations as follows: ˙ ∆xa= Aa· ∆xa+ Ba· ∆u ∆ya= Ca∆xa (20) where ∆xa= ∆x ∆z , ∆ya= ∆y ∆z ∆z= ∆VLdt ∆vdcdt ∆VLdt ∆vdcdt T Aa= A 0 0 C1 0 0 C2 0 0 0 I2×2 0 , Ba= B 0 , Ca= C 0 0 I4×4 (21)
IV. CONTROLDESIGN
The main objective of the proposed controller is to regulate certain output measurements and drive system states to equi-librium operating points when the wind-driven induction gen-erator system is subjected to various disturbances under both grid-connection and islanding conditions. In the design of the output feedback controller, the pole placement approach based on linear quadratic control (LQC) will be used. The goals in the controller design process include both transient excursion suppression and good regulations in steady-state operation. A. Linear Quadratic State Feedback Control
For the linear system described in (20), the linear quadratic state feedback control ∆u that minimizes the performance index
J =1 2 ∞ 0 ∆xTaQ∆xa+ ∆uTR∆u dt (22)
where Q is the weighting matrix of the state variable variations and R that of the control effort, is given by [13]
∆u =−Ks∆xa. (23)
The closed-loop system equation with the optimal control ∆u(t) is ˙ ∆xa= As∆xa, As = Aa− BaKs = Aa− SM (24) where S = BaR−1BTa Ks= R−1BTaM (25)
and Msis the positive semidefinite (stabilizing) solution M of
the algebraic Riccati equation (ARE)
ATaM + M Aa+ Q− MSM = 0. (26)
To determine the state feedback gain matrix Ksfor the linear
quadratic control in (23), a set of initial weights q(0)ij must be first chosen for the weighting matrix Q in (22) based on physical reasoning and past experience. In the present work, a diagonal weighting matrix with the diagonal elements 10, 0, 0, 1, 100, 100, 1, and 1 corresponding to the state variables ∆vdc, ∆ide,
∆iq e, ∆ωur, ∆VLdt, ∆vdcdt, ∆VLdt, and ∆vdcdt,
respectively, is selected as the initial weighting matrix. The weights for the other state variables of minor concern are cho-sen to be zero. Note that the weights for STATCOM currents ∆ide and ∆iq e have been chosen to be small (zero) because it
may be necessary to adjust these variables to new values in order to regulate load bus voltage under various disturbances. Higher but moderate weights are placed on the integral terms∆VLdt,
∆vdcdt to avoid steady-state errors in load bus voltage and
dc-capacitor voltage. However, if an extremely high weight is placed on these variables, the adjustment in the reference cur-rents (∆i∗q eand ∆i∗de) may be limited by the control laws which tries to keep the integral terms fixed at their original values in order to minimize the state variations. With the weighting ma-trices at hand, we can proceed to determine the feedback gain matrix Ksusing (25).
If the dampings for the closed-loop eigenvalues with the opti-mal control in (23) are not satisfactory, the initial weighting ma-trix may be updated using the pole placement technique in [14], where procedures have been described to move the closed-loop eigenvalues by introducing a degree of relative stability χ into the original performance criterion and modifying the weight-ing matrix Q. Then the new spectrum (the collections of stable eigenvalues) Λ of the optimal system with the modified weight-ing matrix (Q− 2χMu) is
Λ = Λs− 2χI (27)
where Mu is the negative semidefinite solution M of (26),
and Λs is the spectrum of the optimal system with the initial
weighting matrix.
To determine Mu, we assume that (Aa, Ba) is stabilizable
and the Hamiltonian matrix
H0= Aa −S −Q −AT a
has no pure imaginary eigenvalues. These conditions are nec-essary and sufficient to ensure the existence of a unique ma-trix Ms = 0 such that (24) is stable. The closed-loop eigen->
structure can be written as Jordan decomposition of H0in the
form [14], [15] Aa −S −Q −AT a X0 W0 Y0 Z0 = X0 W0 Y0 Z0 Λs 0 0 Λu (28) where Λs and Λu are the collections of the stable and
unsta-ble eigenvalues of H0, respectively. Note that Λs and Λu are
symmetric with each other with respect to the imaginary axis. The matrix Ms can be derived from the eigenvectors of H0
associated with the stable eigenvalues in (28) in the form
Ms = Y0X−10 . (29)
Similarly, a negative semidefinite matrix
Mu = Z0W−10 (30)
exists and is also a solution of the ARE in (26) associated with the unstable eigenvalues.
The results of [14] allow a desired shift in the locations of the original eigenvalues by using a negative semidefinite ma-trix (Mu) that is obtained from the initial weighting matrix. It
is noted that partial placement of optimal eigenvalues can be reached in the same manner and will be used in this paper. B. Output Feedback Control
Since some state variables such as ∆φq r, ∆φdr, ∆φq s, and
∆φds are not readily measurable, nine output variables ∆VL,
∆vdc, ∆ωur, ∆iq e, ∆ide, ∆VLdt, ∆vdcdt, ∆VLdt,
and ∆vdcdt as listed in (21) are selected to be the feedback
signals. Then, the output feedback control ∆u is given by
TABLE I
SUMMARY OFSYSTEMEIGENVALUES
where the output feedback gain matrix Kois related to the state
feedback gain matrix Ksas follows:
KoCa= Ks. (32)
Thus, the output feedback gain matrix Kocan be derived using Ksand C†a, the pseudoinverse [9], [16] of the augmented output
matrix Ca, as follows: Ko= KsC†a (33) where C†a= CTa CaCTa −1 .
Sets of open-loop and closed-loop eigenvalues for the WECS are summarized in Table I. The relative stability of χ = 6 is in-troduced to shift the eigenvalues for the electromechanical mode from−5.79 ± j7.94 to −17.79 ± j7.91 in order to improve the damping effect for the closed-loop system with state feedback controller. On the other hand, the eigenvalues for the electrome-chanical mode are shifted to new locations −9.61 ± j6.17 if an output feedback controller with generator speed feedback and pitch angle control is used. In this case, the diagonal elements of the modified state weighting matrix correspond-ing to ∆vdc, ∆ide, ∆iq e, ∆ωru,
∆VL, ∆vdc, ∆VL, and ∆vdcbecome 9.46, 0.00018, 0.00024, 53221, 207.65,
557.77, 197.88, and 205.28, respectively. It is clear that the modified weight on ∆ωruis higher than that of the initial weight
since improved damping for the electromechanical mode is de-sired. In addition, in the grid-connection condition, a poorly damped STATCOM mode characterized by the pair of eigen-values−45.68 ± j376.44 is present in the open-loop response. This mode can be shifted by the output feedback controller to the new locations−97.29 ± j553.17. Moreover, the pair of well-damped modes for controller dynamics in Table I indicate that the deviations of load bus voltage and dc-capacitor voltage would be reduced by the proposed output feedback controller.
Fig. 5. Dynamic responses for the WECS operated in grid-connection condi-tion. (A 10-cycle temporary voltage dip of 5% at grid began at t1= 0.095 s).
(Wind gust began at t2= 3.0 s). (Wind gust ended at t3= 6.0 s). (A step load
change of 10% began at t4= 8.0 s).
Fig. 6. Dynamic responses for load bus voltage when the WECS was operated in grid-connection condition. (A temporary voltage dip of 5% at grid began at
t1= 0.095 s).
V. SIMULATIONRESULTS A. Grid-Connection Performance
The dynamic performances of the output feedback controller for the WECS were examined under three successive distur-bances and the results are depicted in Figs. 5–8. The first event was a temporary voltage dip of 5% at the grid bus starting at t1= 0.095 s and lasting for ten cycles. The voltage dip caused
a transient swing of±0.2 rad/s in generator speed, as shown in Fig. 5(c). The response curves in Fig. 5(c) and (d) reveal that the blade pitch angle can be controlled to limit generator speed deviation in this event.
The second disturbance of a wind gust started at t2= 3.0 s,
reaching the peak wind speed of 46 mph at t = 4.6 s, and fi-nally the wind speed dropped to 36 mph at t3= 6.0 s. For the
Fig. 7. Dynamic responses for load bus voltage when the WECS was operated in grid-connection condition. (Wind gust began at t2= 3.0 s). (Wind gust ended
at t3= 6.0 s).
Fig. 8. Dynamic responses for load bus voltage when the WECS was operated in grid-connection condition. (A step load change of 10% began at t4= 8.0 s).
sake of clarity, the effects of wind shear and tower shadow were ignored. It is obvious that the rotor would accelerate as a result of the increasing mechanical power caused by increasing wind speed if the pitch angle was fixed. However, as demonstrated in Fig. 5(d), the wind-sensorless output feedback controller gen-erated a pitch angle command as soon as the wind speed be-gan to increase. In addition, the increased output power con-verted from aerodynamic power after gust was another benefit of the controller, as illustrated in Fig. 5(b). The third distur-bance considered was a 10% step change in load demand that took place at t4= 8.0 s. This sudden increase in load demand
caused a transient drop in the kinetic energy and speed of the IG in order to provide the sudden increase in generator output power, as shown in Fig. 5(b) and (c). After this transient, the increase in load power was balanced by the grid, and generator speed returned to its original value as shown in Fig. 5(c). It is also observed from Fig. 5(d) that the pitch angle decreased slightly during the transient and returned to the original value afterward.
An observation of the response curve in Fig. 5(a) indicates that the dc-capacitor voltage can be kept almost constant under these disturbances except for a very small deviation of±0.4% in the voltage dip event.
It is also observed from the response curves in Figs. 6–8 that the STATCOM with the output feedback controller is su-perior to the fixed shunt capacitor in providing voltage support for the WECS under large disturbances. As illustrated in Fig. 6, the load bus voltage supported by STATCOM settled to the steady-state value in 2 cycles after the sudden change in grid voltage. Besides, it is clear that a very small transient excur-sion of±0.015% in load bus voltage was observed during the wind gust transient, as shown in Fig. 7. A transient excursion of less than 0.3% and lasting for only 2 cycles is observed for the load bus voltage in Fig. 8 when the system was subjected to a step load change. In contrast to the STATCOM, the voltage supported by the fixed shunt capacitor experienced considerably higher voltage excursions and steady-state errors, as shown in Figs. 6–8.
Fig. 9. Dynamic responses for the WECS operated in islanding condition. (Grid disconnection began at t1= 0.095 s). (Wind gust began at t2= 3.0 s).
(Wind gust ended at t3= 6.0 s). (A step load change of 10% began at t4=
8.0 s).
B. Islanding Performance
Self-excitation can occur if the IG loses its connection to the grid. Unstable voltage and frequency may be observed during self-excitation [7], [8]. In the first simulated event, the WECS lost its connection to the grid at t1= 0.095 s and remained in
islanding mode thereafter. Fig. 9(b) illustrates that the output power of IG dropped rapidly from 0.798 to 0.488 p.u. after the grid-disconnection transient since the IG exported a real power of 0.31 p.u. to the grid before it was disconnected from the grid. Note that the load demand remained unchanged when the IG was disconnected from the grid since the load bus voltage mag-nitude was fixed by the STATCOM at steady state. This is as expected since constant impedance load model was employed in this simulation. During the grid-disconnection transient, the surplus mechanical power resulting from the decrease in electri-cal power output was conditioned by increasing the blade pitch angle β. Generator speed can thus be stabilized to a new value, as shown in Fig. 9(c) and (d). The frequency difference ∆fo
(Hz) between the grid and the islanding induction generator is depicted as the rate of change in angular displacement (dθL/dt,
which is equal to 2π∆foin rad/s) in Fig. 9(e). It is observed from
Fig. 3 that the θL, the relative angular displacement between the
Fig. 10. Dynamic responses for load bus voltage when the WECS was operated in islanding condition. (Grid disconnection began at t1= 0.095 s).
Fig. 11. Dynamic responses for load bus voltage when the WECS was operated in islanding condition. (Wind gust began at t2= 3.0 s). (Wind gust ended at
t3= 6.0 s).
Fig. 12. Dynamic responses for load bus voltage when the WECS was operated in islanding condition. (A step load change of 10% began at t4= 8.0 s).
can not be fixed since the load bus voltage vector VL rotates
at a speed ω that is different from the synchronous speed ωs
for the grid voltage vector V∞under islanding condition. Thus dθL/dt no longer equaled zero and there was a frequency
dif-ference between the islanding IG and the grid.
When the WECS was operated under islanding condition, a wind gust and a step load change of 10% were assumed to take place at t2= 3.0 s and t4= 8.0 s, respectively, in order to
examine the dynamic performance of the output feedback con-troller under disturbance conditions. It is observed from Fig. 9(c) and (e) that both generator speed (ωr) and the rate of change
of relative angular displacement (dθL/dt) can be stabilized to
new equilibrium points by the control of blade pitch angle β [Fig. 9(d)] under islanding condition.
As shown in Fig. 9(a), the dc-capacitor voltage can be kept al-most constant under all disturbances except for a small transient excursion of 1.8% during the grid-disconnection event.
Figs. 10–12 summarize the voltage regulation performances of output feedback controller for the WECS in islanding con-dition. As illustrated in Fig. 10, the load bus voltage settled to a steady-state value in 10 cycles during the grid-disconnection transient. Figs. 11 and 12 show that the transient excursions in load bus voltage can be suppressed to be less than±0.02% and±1% by the controller for the disturbances of wind gust and sudden change in load, respectively. Besides, undervoltage (underexcitation) may occur due to insufficient reactive power compensation by fixed shunt capacitor for the IG when it is disconnected from the grid, as illustrated in Fig. 10.
Fig. 13. Induction generator input–output variables for steady-state analysis in islanding.
VI. STEADY-STATEANALYSIS FORFREQUENCYCONTROL OF ANISOLATEDINDUCTIONGENERATOR
It has been observed from the response curve in Fig. 9(e) that system frequency no longer remained fixed at the rated value of 60 Hz when the induction generator was disconnected from the power grid and was operated under islanding condition. In fact, frequency deviations of 0.485 Hz (3.051 rad/s) and 0.454 Hz (2.850 rad/s) were observed for the isolated generator when it was subjected to wind gust and a step load change, respectively. Under the assumption of sufficient mechanical power input, it is interesting to study how to reduce the frequency deviations and maintain constant frequency operation for an isolated in-duction generator. To do this, the relationship between rotating speed (ωr) and real power output (Pg) at various frequencies
for the isolated induction generator under steady-state operating condition must be first determined.
A. Algorithm for Determining the Steady-State Operating Points of an Isolated Induction Generator
To find the steady-state operating points for an isolated in-duction generator with a shunt excitation capacitor, an iterative approach has been proposed [3] to solve the nodal equations for the induction generator system. However, it is difficult to formulate nodal equations for an IG with a voltage regulator. In this paper, a simple steady-state analysis algorithm based on dynamic system model is proposed for an isolated IG with a voltage regulator using the MATLAB function TRIM. The volt-age regulator can be a STATCOM, an SVC, or a capacitor with adjustable capacity. As shown in Fig. 13, the constrained in-puts to the induction generator system include load bus voltage VL, generator output Pg, and system frequency fo. The
uncon-strained outputs include mechanical power Pm, generator speed
ωr, and the required reactive power Qe. Fig. 14 depicts the
flow chart for the proposed steady-state analysis algorithm. In the algorithm, proper capacity of the equivalent shunt capacitor can be determined such that the load bus voltage VL is fixed at
0.986 p.u. The generator speed ωris computed for various
com-binations of electrical output power Pg and system frequency
fousing the power repetition loop and the frequency repetition
loop in Fig. 14. Note that the equivalent load impedance must be updated in the power repetition loop when the constrained gen-erator output power is changed. In addition, the angular speed ωs for the synchronous reference frame and the base quantity
ωb must be set to be equal to the specified frequency 2πfo in
Fig. 14. Flow chart for the proposed steady-state analysis algorithm for an IG operated in islanding.
constant HT and damping constant DT must also be updated
based on the specified frequency 2πfo.
Since the nonlinear induction generator model has more than one steady-state solution, the reasonable steady-state solution hinges on the initial value. To acquire reasonable steady-state solution and speed up convergence while using the proposed algorithm, the first initial condition can be obtained by solving the dynamic equations of grid-connected IG since the output frequency is the same as the grid frequency at steady-state, irrespective of generator speed or load power, thus confining the solution to a small region. In addition, steady-state solution in each power repetition loop can be treated as the initial value for the next power repetition loop to speed up convergence, as shown in Fig. 14.
Fig. 15 shows the results from steady-state analysis for output frequencies foranging from 59.4 to 60.6 in 0.2-Hz increment
with the load bus voltage fixed at 0.986 p.u. It is observed from Fig. 15 that generator speed ωrincreases linearly with increasing
real power output Pg at a specific output frequency. It is also
observed that a change in output frequency means a change in rotor speed when the real power output is held constant.
Note that the steady-state operating points at ta= 2.5 s,
tb = 7.5 s, and tc= 9.5 s in Fig. 9 are marked in Fig. 15.
The corresponding frequency deviations for the three operating points in Fig. 15 are 0.272 Hz (1.713 rad/s), 0.485 Hz (3.051 rad/s), and 0.454 Hz (2.850 rad/s), respectively, which match closely with the dynamic simulation results in Fig. 9(e).
Fig. 15. Rotating speed (ωr) versus real power output (Pg) at different
fre-quencies for an isolated IG.
Fig. 16. Slope variation of Pg versus ωr curve with various output
frequen-cies.
B. Constant Frequency Controller Design
Fig. 15 gives the relationship between generator speed ωr,
frequency fo, and real power output Pg for an islanding IG
under constant voltage control. An interesting application of the steady-state operating characteristics in Fig. 15 would be the constant frequency operating mode for an induction generator with sufficient mechanical power Pm. In the design of a constant
frequency controller for such an isolated IG, the generator speed ωr in Fig. 15 is first expressed as a linear function of its real
power output Pg:
ωr|fo = Ω(Pg) = aPg + b (rad/s). (34)
Since the IG delivers zero output power when it is operated at synchronous rotating speed (ωr = 2πfo), the parameter b in
(34) equals the synchronous frequency 2πfo. In addition, the
slope a in (34) at each frequency focan be measured from the
curves in Fig. 15 and is plotted in Fig. 16. It is observed from Fig. 16 that the slope a can be approximated as a linear function of foas follows:
a(fo) = 0.02125fo+ 0.0071. (35)
Substituting (35) into (34) yields the general form for the ref-erence rotor speed in terms of the actual real-power output and the specified output frequency as
ωr∗= Ω∗(Pg, fo∗) = a(fo∗)Pg+ 2πfo∗ (rad/s). (36)
Based on (36), a proportional-integral controller as shown in Fig. 17 is proposed to reach constant frequency control for an isolated induction generator. As shown in Fig. 17, the inputs to the proposed constant frequency controller are the desired frequency fo∗(fo∗= 60 Hz in this work), the real power output
Fig. 17. Constant frequency controller for an IG operated in islanding condi-tion.
Fig. 18. Dynamic responses for the wind-driven IG with constant frequency controller and without constant frequency controller. (Grid disconnection began at t1= 0.095 s).
Pg, and the generator speed ωr. The reference speed ω∗ris
deter-mined using (36). When the actual speed ωr deviates from the
reference speed ω∗r(e.g., ωr>ω∗r), then the PI controller
gener-ates the desired pitch angle command ∆βC F based on the speed
deviation (∆ω = ω∗r− ωr). A negative ∆ω causes an increase
in pitch angle (∆βC F>0). As a result, the mechanical power
input to the generator is decreased, causing a decrease in the kinetic energy and rotating speed of the generator. Finally, the rotating speed ωrsettles down to a lower steady-state value and
the frequency deviation is eliminated.
To examine the effectiveness of the proposed constant fre-quency controller, dynamic responses for the induction gener-ator that was disconnected from the grid at t1= 0.095 s are
depicted in Fig. 18. It is observed from Fig. 18(d) that zero-frequency deviation (dθL/dt = 2π∆fo= 0) can be reached by
the proposed constant frequency controller. This is achieved by increasing the pitch angle β, causing a temporary decrease in mechanical power Pm and a decrease in generator speed ωr, as
shown in Fig. 18(a)–(c). For comparison, the generator speed ωr for the output feedback without constant frequency control
is also shown in Fig. 18(b). It can be observed that the output feedback controller without constant frequency control can pro-vide stable operation with satisfactory response under islanding operation except for a small frequency deviation as shown in Fig. 18(d). This means that the mechanical power Pm, during
the grid-disconnection transient, was controlled to be slightly higher than the real-power output Pg so that the accumulated
kinetic energy accelerated IG to a higher rotating speed, as il-lustrated in Fig. 18(a) and (b).
Another important application of (36) is the design of a con-stant output power controller for the IG connected with a grid. In this case, the generator speed reference ωr∗is derived based on the desired real power output Pg∗ and system frequency fo,
as shown in Fig. 17.
VII. CONCLUSION
A STATCOM and blade pitch control system has been suc-cessfully designed to regulate the load bus voltage and stabi-lize rotor speed for an induction generator in a variable-speed WECS. A major feature of the present work is that the two-reference-frame model is included in the detailed dynamic sys-tem model in order to achieve decoupled control for the STAT-COM real and reactive power control loops. Besides, a sys-tematic approach based on pole placement technique has been developed to determine proper weighting matrix for the linear quadratic controller such that the eigenvalues for the closed-loop system can be shifted to desired locations.
To ensure zero steady-state error regulation and improve transient excursion for load bus voltage and dc-capacitor volt-age, four new states, ∆VLdt,
∆vdcdt,
∆VLdt, and
∆vdcdt, have been augmented in the system dynamic model.
Simulation results indicate that the output feedback controller designed for the STATCOM and blade pitch is capable of im-proving WECS dynamic responses in case of disturbances un-der grid-connection or islanding conditions. It is also noted that generator speed feedback instead of wind speed feedback for the output feedback controller provides satisfactory transient response in stabilizing generator speed. A simple algorithm based on dynamic system model for an isolated IG has been developed such that a flexible steady-state analysis is achiev-able under various operating conditions. Utilizing the steady-state properties deduced by means of the proposed algorithm, a proportional-integral controller has been designed such that constant frequency operation (for an isolated IG) and constant-output power operation (for a grid-connected IG) can both be reached as long as the mechanical power is sufficient. The pro-posed constant output frequency controller and constant output power controller are also suitable for applications in hydro-driven or steam-hydro-driven induction generator systems.
APPENDIX 1) Fixed Shunt Capacitor Model:
˙vdL = ωsvq L+ ωbXcidFC
2) Transmission Line Model: ˙idTL=− ωbrTL XTL idTL+ ωsiq TL+ ωb XTL (vdL − vd∞) ˙iq TL=− ωbrTL XTL iq TL− ωsidTL+ ωb XTL (vq L− vq∞). 3) Load Model: ˙idL =− ωbrL XL idL+ ωsiq L+ ωb XL vdL ˙iq L =− ωbrL XL iq L− ωsidL+ ωb XL vq L. 4) Simulation Parameters: REFERENCES
[1] A. Murdoch, R. S. Barton, J. R. Winkelman, and S. H. Javid, “Control design and performance analysis of a 6 MW wind turbine-generator,”
IEEE Trans. Power App. Syst., vol. PAS-102, no. 5, pp. 1340–1347, Dec.
1983.
[2] L. L. Freris, Wind Energy Conversion Systems. Englewood Cliffs, NJ: Prentice-Hall, 1990.
[3] S. S. Murthy, O. P. Malik, and A. K. Tandon, “Analysis of self-excited induction generators,” Proc. Inst. Elect. Eng.–C, vol. 129, no. 6, pp. 260– 265, Nov. 1982.
[4] C. D. Schauder and H. Mehta, “Vector analysis and control of advanced static VAR compensators,” Proc. Inst. Elect. Eng.–C, vol. 140, no. 4, pp. 299–306, Jul. 1993.
[5] C. T. Chang and Y. Y. Hsu, “Design of UPFC controllers and supple-mentary damping controller for power transmission control and stability enhancement of a longitudinal power system,” Proc. IEE–Gener. Transm.
Distrib., vol. 149, no. 4, pp. 463–471, Jul. 2002.
[6] W. L. Chen and Y. Y. Hsu, “Direct output voltage control of a static syn-chronous compensator using current sensorless d–q vector-based power balancing scheme,” in Proc. IEEE/PES T&D Summer Meet., 2003, vol. 2, pp. 545–549.
[7] Z. Saad-Saoud, M. L. Lisboa, J. B. Ekanayake, N. Jenkins, and G. Strbac, “Application of STATCOMs to wind farms,” Proc. IEE–Gener. Transm.
Distrib., vol. 145, no. 5, pp. 511–516, Sep. 1998.
[8] W. L. Chen and Y. Y. Hsu, “A novel active voltage and frequency reg-ulator to improve grid-disconnection transients of self-excited induction generator system,” in Proc. 5th IEEE Int. Conf. Power Electron. Drive
Syst., 2003, vol. 2, pp. 1586–1590.
[9] E. S. Abdin and W. Xu, “Control design and dynamic performance analysis of a wind turbine-induction generator unit,” IEEE Trans. Energy Convers., vol. 15, no. 1, pp. 91–96, Mar. 2000.
[10] C. M. Ong, Dynamic Simulation of Electric Machinery Using
MAT-LAB/SIMULINK. Englewood Cliffs, NJ: Prentice-Hall, 1998. [11] P. M. Anderson and A. Bose, “Stability simulation of wind turbine
sys-tems,” IEEE Trans. Power App. Syst., vol. PAS-102, no. 12, pp. 3791– 3795, Dec. 1983.
[12] O. Wasynczuk, D. T. Man, and J. P. Sullivan, “Dynamic behavior of a class of wind turbine generators during random wind fluctuations,” IEEE
Trans. Power App. Syst., vol. 100, no. 6, pp. 2837–2845, Jun. 1981.
[13] F. L. Lewis and V. L. Syrmos, Optimal Control. New York: Wiley, 1995. [14] J. Medanic, H. S. Tharp, and W. R. Perkins, “Pole placement by per-formance criterion modification,” IEEE Trans. Autom. Control, vol. 33, no. 5, pp. 469–472, May 1988.
[15] J. C. Willems, “Least squares stationary optimal control and the algebraic Riccati equation,” IEEE Trans. Autom. Control, vol. 16, no. 6, pp. 621– 634, Dec. 1971.
[16] P. H. Huang and Y. Y. Hsu, “An output feedback controller for a syn-chronous generator,” IEEE Trans. Aerosp. Electron. Syst., vol. 26, no. 2, pp. 337–344, Mar. 1990.
Woei-Luen Chen (S’03–M’04) was born in Taiwan, R.O.C., on November 29,
1971. He received the M.S.E.E. degree from the National Taiwan University in 1997. Currently he is a Ph.D. candidate at the Department of Electrical Engineering, NTU.
His current research interests include power electronic applications, reactive power compensation systems, and the wind-energy conversion systems.
Yuan-Yih Hsu (S’82–M’82–SM’89) was born in Taiwan, R.O.C., on June 19,
1955. His current research interests include power system dynamics and stability analysis, distribution automation, and the application of artificial intelligence to power systems.
Since 1977, he has been with the National Taiwan University, where he is now a Professor.
