Sunteți pe pagina 1din 26

INTERNATIONAL TRANSACTIONS ON ELECTRICAL ENERGY SYSTEMS

Int. Trans. Electr. Energ. Syst. (2015)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/etep.2120

Optimum design of proportional-integral controllers in grid-


integrated PMSG-based wind energy conversion system

Saurabh M. Tripathi2*,, Amar Nath Tiwari1 and Deependra Singh2


1
Department of Electrical Engineering, Madan Mohan Malaviya University of Technology, Gorakhpur, (U.P.), India
2
Department of Electrical Engineering, Kamla Nehru Institute of Technology, Sultanpur, (U.P.), India

SUMMARY
In this paper, a direct-driven small permanent magnet synchronous generator (PMSG) wind energy con-
version system (WECS) with a back-to-back power converter working in grid-connected mode is
discussed. Control structures based on eld-oriented control and voltage-oriented control mechanisms
are proposed for machine-side converter and grid-side converter, respectively. Optimization approaches
namely modulus optimum and symmetric optimum are used to obtain analytical expressions for
the selection of the parameters of involved proportional-integral (PI) controllers in different control
loops. A transient system simulation using SimPowerSystem is built to evaluate the performance of
the PMSG-based WECS by employing selected values of PI controller parameters both under varying
wind conditions and under symmetrical and asymmetrical grid-fault conditions. Copyright 2015 John
Wiley & Sons, Ltd.

key words: Fault ride-through (FRT) eld-oriented control (FOC); grid-side converter (GSC); machine-
side converter (MSC); modulus optimum (MO); permanent magnet synchronous generator
(PMSG); symmetric optimum (SO); voltage-oriented control (VOC); wind energy
conversion system (WECS)

1. INTRODUCTION

With the recent advances in power converter technology and availability of permanent magnet mate-
rials, direct-drive permanent magnet synchronous generators (PMSGs) have increasingly drawn the at-
tention of wind turbine manufacturers [1]. PMSG offers high efciency, low maintenance, no
excitation system, and improved power factor and elimination of the gearbox [26]. Moreover, a
full-scale power electronic converter between the PMSG and the utility grid offers not only additional
technical performances such as complete decoupling from the grid, full controllability of the system for
maximum wind power extraction, high performance, high precision, high reliability, wide operating
range and improved capability of fault-ride-through (FRT) but also paybacks in the losses occurring
in the power conversion stage [1,7]. Pulse width modulation (PWM) voltage source converter
(VSC) is the state-of-the-art technology nowadays increasingly favored by all wind turbine manufac-
tures. The possibility of high switching frequencies accompanied by an appropriate control strategy
makes PWM converters t for grid interface of wind generation system [7].
Most of the industrial controllers still in operation are based on PI control law, which offers the sim-
plest, feasible, and most efcient solution to control problems [810]. Parametric tuning and optimiza-
tion of the PI controller can be viewed as one of the most crucial engineering task during the
commissioning of control system so as to obtain the desired control responses [11]. Parametric tuning
of the PI controller is a compromise between speed of response and stability for small-signal distur-
bances as well as robustness to tolerate large-signal disturbances [12]. The relative stability of a control

*Correspondence to: Saurabh M. Tripathi, Department of Electrical Engineering, Kamla Nehru Institute of Technology,
Sultanpur228 118 (U.P.), India.

E-mail: mani_excel@yahoo.co.in

Copyright 2015 John Wiley & Sons, Ltd.


S. M. TRIPATHI ET AL.

system can be analyzed using two quantitative measures dened as gain margin (GM) and phase
margin (PM) determined from the Bode plot of open-loop transfer function. Further, the locations
of the closed-loop poles decide the transient performance and stability of a control system.
A large amount of research work has already been carried out in the past to determine the PI
controllers parameters [13]. Numerous tuning techniques are summarized in [11,14] and the refer-
ences therein. Each tuning technique has its own features and constraints. Brief descriptions of a
few PI tuning and optimization techniques are presented in the following paragraphs along with their
key features and limitations so as to stimulate the readers to compare and judge each of these tech-
niques individually for their superiority.
One of the most popular Zeigler-Nichols (Z-N) method offers a rough estimate of the PI controller
parameters, which need to be further adjusted heuristically by the designers so as to achieve the desired
closed-loop response and is particularly suitable for the system with monotonic step response (S-shape
response) [13,15]. The PI controller tuned according to Z-N method may cause high overshoot, large
oscillation, and longer settling time in closed-loop response of a higher-order system [13]. The
approach based on internal model control is another most commonly used PI tuning technique, which
essentially requires the plant model to be reduced to rst or second order before controller design rules
could be applied [16]. However, the choice of the reduction method greatly inuences the resulting
controller design [16].
In order to address different aspects of the controller design problem, several methods employ
optimization techniques. For instance, [17] copes with controller optimization with respect to the
criteria ISE (integral-square error), IAE (integral-absolute error), and ITAE (integral time absolute
error). By focusing on the square of the error function, the ISE criterion penalizes the positive as well
as negative values of the errors, whereas by focusing on the magnitude of the error function, the IAE
criterion penalizes either the positive or negative values of the errors [18]. Because both ISE and IAE
indices weight all errors equally independent of time, their minimization may result in step response
with relatively small overshoot but a longer settling time [19,20]. The long duration transients in step
response can be penalized by proposing the criteria ITSE (integral time square error) and ITAE
[18,19]. Although the ITSE and ITAE criteria may overcome the disadvantages associated with the
ISE and IAE criteria, respectively, the process of deriving the analytical formula becomes complex
and time-consuming [20].
Mostly, PI tuning techniques are model based, and therefore, the PI controller might not lead to
specied control performance if the controlled plant is uncertain [10,21]. This is why the designers
always look for robust control design by synthesizing a controller for which the closed-loop system
is stable and the specied control performance could be achieved despite plant uncertainty [21,22].
In fact, the systems robustness (i.e., the systems ability to withstand changes in its parameters before
becoming unstable) is highly reliant on the greatness of GM and PM [21]. Intensive researches have
been published by many authors on designing PI controllers so as to meet GM and PM specications.
Besides robustness, a good closed-loop control performance is also vital, and therefore, a trade-off
between the robustness and the control performance is usually taken into consideration while synthe-
sizing the PI/PID controller parameters [21]. For instance, a PID controller has been designed in [21]
based on non-linear optimization wherein the closed-loop bandwidth is maximized for specied GM
and PM with constraint on overshoot ratio so as to satisfy the criteria of both robustness and closed-
loop control performance in the design.
Genetic algorithm (GA)-based methods have recently received great interest in searching global op-
timal solution in PI controller design problem [20]. However, the natural genetic operations of GA-
based methods often lead to enormous computational efforts compared with other traditional methods
to give an optimal solution [13,20]. Moreover, the premature convergence of GA algorithm not only
degrades its performance but also reduces its search capability [13,20]. Further, some deciencies in
GA performance are also apparent in applications where the parameters being optimized are highly
correlated [20]. The particle swarm optimization (PSO) is one of the modern heuristic optimization al-
gorithms, which can circumvent the premature convergence and can generate high-quality solution to a
PI controller design problem with no complicated evolutionary operations (i.e., selection, crossover,
and mutation), and therefore, it reduces the computational efforts to some extent compared with the
GA method [10,20]. Nevertheless, the PSO is usually employed as a minor compensatory tuner [10].

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

Loop-shaping criteria viz. modulus optimum (MO) and symmetric optimum (SO) are two fairly sim-
ple and generic PI controller tuning techniques for designing optimal linear-control system in the fre-
quency domain when certain parameters of the system are known [2325]. These criteria are
introduced in [9,11,12,25,26] and have conventionally been used for the evaluation of involved PI con-
troller parameters in vector control structures of electric motor drives [12,23,27]. An experimental per-
formance evaluation of PI tuning techniques based on MO and SO criteria applied to the eld-oriented
control structure of a permanent magnet synchronous motor drive has been presented in [28]. A
discrete-time equivalent to MO and SO criteria has been investigated in [23] and applied to the vector
control of grid-connected VSC. MO and SO prove to be a feasible controller design technique with a
simple extension to apply it even on higher-order plants using the active damping approach proposed
in [24]. To design PID controller for type-p control loops, which is characterized by the presence of
p integrators in the open-loop transfer function, the SO criterion has been extended in [9]. The MO
and SO criteria present at least two major advantages in the design of control loops(i) complete plant
model is not a requisite and (ii) set-point response of the closed-loop system is satisfactory [9].
Among different techniques of PI controller parameters tuning, which differ in complexity, exibil-
ity, and requirement of plant knowledge, there is always a need for a simple, straightforward, and in-
tuitive technique requiring least plant knowledge and offering the best possible control performances
[11,21]. Despite the widespread usage of PI controllers in the control structures of WECSs employing
different sorts of generator technology, there are only a very few literatures that address the optimum
design of controller parameters in wind power applications. For instance, the authors of [29] presented
an optimum coordinated controller design for PMSG-based WECS where controller parameters are de-
termined by optimizing the considered performance indices such as peak overshoot and settling time.
It has been realized that the optimum design of PI controllers in the control structures of a WECS so
as to obtain the best possible trade-off between closed-loop control performance and robustness is still
a topic that is regarded to be poorly reported, and this paper addresses this need by presenting a sys-
tematic design procedure for a grid-connected WECS employing small PMSG along with the sizing of
components and evaluation of involved PI controllers in different control loops of the same using eas-
ily understandable MO and SO criteria. The problem is formulated as Tune the involved PI controller
parameters in the control structures of a PMSG-based WECS such that the best possible trade-off be-
tween closed-loop control performance and systems robustness margin could be realized.

2. SYSTEM DESCRIPTION AND CONTROL PHILOSOPHY

The proposed WECS mainly consists of a wind turbine, a PMSG with surface-mounted permanent
magnets, a frequency converter built by two current regulated PWM VSCs namely (i) machine-side
converter (MSC) and (ii) grid-side converter (GSC), and a common dc-link capacitor in between as
shown in Figure 1.

2.1. Wind turbine model


The output power of the wind turbine [30] is expressed as
Pm 0:5C p AV 3 (1)
where Pm, , Cp, A, R, and V are the turbine output power, air density, power coefcient, swept area
(= R2), radius of the turbine blades and wind speed, respectively.
The power coefcient Cp is a function of the tip-speed ratio and pitch angle [31], expressed as
   
151 18:4
C p ; 0:73  0:58  0:0022:14  13:2 exp  (2)
i i
1 1 0:035
 (3)
i 0:08 3 1
r R=V (4)
where r is the turbine rotational speed, and for lower to medium wind speeds, can be set to zero [32].

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

Figure 1. Proposed PMSG-based wind energy conversion system.

2.2. Generator model


The stator voltage equations of PMSG in d-q reference frame [33,34] are expressed as
d
vds Rs ids Ld ids  s Lq iqs (5)
dt
d
vqs Rs iqs Lq iqs s Ld ids s m (6)
dt
where ids and iqs are the d - q axes stator currents, Rs is the stator resistance, Ld and Lq are the d - q axes
stator inductances, m is the rotor ux, and s is the electrical speed. The electromagnetic torque [34] is
given as
3 P   
T e   m iqs Ld  Lq ids iqs (7)
2 2
where P is the number of PMSG poles. It is noticeable that for low-speed PMSG with surface-mounted
permanent magnets, d and q axes inductances are the same; i.e., Ld Lq [32]. The parameters of PMSG
[35] are listed in the appendix.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

2.3. DC-link capacitor


A dc-link capacitor is essential to lter out dc voltage ripples across it [36]. For satisfactory PWM con-
trol, the dc-link voltage VDC is selected [31] as follows
p p
V DC > 2 2V LL = ma 3 (8)
where VLL is the line-line root-mean-square (r.m.s.) voltage on the ac side of the PWM converter and
ma is the modulation index considered equal to maximum value of one. It is to be noted that the con-
straint (8) on VDC is from the r.m.s. voltages on the ac sides of both MSC and GSC.
The dc-link capacitor value is estimated [31] as
p
C 0:9I peak = 4 2f V DCripple (9)
where Ipeak is the permissible peak ac line current, f is the grid frequency, and V DCripple is the permissible
ripple dc-link voltage (2%).

2.4. Coupling AC inductor


The value of coupling ac inductor in between GSC and the grid is calculated [31] as
p  
Lg 3ma V DC = 12s f gsw ip (10)
where s is safety factor (120%), fgsw is the switching frequency of the GSC (10 kHz selected), and ip is
the peak-to-peak permissible ripple current (5%).

2.5. Maximum power point tracking


In view of the fact that any amount of power generated by the wind generation system can be injected
into the grid, a grid-integrated WECS is always operated at maximum power point (MPP) to maximize
the generation and utilization of power [1,37]. For maximum power point tracking (MPPT), the PMSG
is operated in variablespeedvariablefrequency mode, where the rotor speed is allowed to vary in
sympathy with the wind speed by maintaining to the optimum value [38]. The wind speed is mea-
sured using an anemometer, and a reference rotor speed corresponding to the MPP is generated using
Equation (11).
r opt V =R (11)

2.6. Controller blocks


Controller blocks are chosen to be PI controllers for different control loops. Small overshoot, good
damping of oscillations, and fast response are the three fundamental goals of the designers for the syn-
thesis of involved PI controller in a control loop. The evaluation of PI controller parameters is one of
the key issues in the design of a cascaded control structure where inner loops are designed to achieve
fast response and outer loop is designed to achieve optimum regulation and stability [12]. As this paper
deals with the synthesis of involved PI controllers in different control loops of a PMSG-based WECS
using MO/SO criteria, brief ideas of these criteria clearly stating the objectives of their formulations are
presented in the following subsections.

2.6.1. Modulus optimum. The MO is an optimization criterion generally dealing with the PI controller
design for second-order control systems having one dominant time constant and other minor time con-
stant [24,39]. The MO is formulated with the design objective to obtain a controller that gives the mag-
nitude of the frequency response of the closed-loop transfer function as at and as close to unity for a
large bandwidth as possible [11,25,40]. If G(p) is the closed-loop transfer function from the set point to
d n j Gj j
the output, the controller is determined in such a way that G(0) = 1 and dn 0 for as many n
0
as possible [11].

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

This method is used because of its simplicity and fast response. The standard form of the control
systems transfer function for the MO is achieved by canceling the dominant time constant, and there-
fore, the systems performance in response to any disturbance applied anywhere other than the set
point is not optimal [25]. The MO results in a fast and non-oscillatory closed-loop time response for
a large class of plants [25]. When it is possible to approximate the plant model with the benchmark
transfer function of the form
K1
GP1 p ; T 1 >> T 2 (12)
1 pT 1 1 pT 2
where K1 is the plant gain, T1 is the plants dominant time constant, and T2 is the plants parasitic time
constant or minor time constant; the MO criterion leads to the proportional gain Kp1,MO and integral
time constant Ti1,MO of the PI controller expressed as

T1
K p1; MO and T i1; MO T 2 (13)
2 T 2K1

2.6.2. Symmetric optimum. When one pole of the open-loop transfer function is near to the origin or at
the origin itself, the SO criterion is used for the evaluation of the PI controller parameters [12]. Sug-
gested as an extension of the MO, the SO criterion is formulated with the design objective to obtain
a controller that compels the magnitude of the closed-loop frequency response as close to unity in
the widest possible frequency range as possible and maximizes the phase margin for given frequency
so that the system can tolerate delays and, in addition, optimizes the behavior of the control system
with respect to the disturbances affecting the control plant [9,12,24,26]. The systems nonlinearities
and time-varying parameters are also well-handled by the PI controller designed using the SO crite-
rion. When it is possible to approximate the plant model with the benchmark transfer function of the
form given by Equation (12), the SO criterion leads to the proportional gain Kp1,SO and integral time
constant Ti1,SO of the PI controller expressed as
T1
K p1; SO and T i1; SO a2 T 2 (14)
a T 2K 1
Instead, when it is possible to approximate the plant model with the benchmark transfer function of the
form given by
K2
GP2 p (15)
p 1 pT 2
where K2 is the plant gain and T2 is the plants parasitic time constant; the SO criterion leads to the
proportional gain Kp2,SO, and integral time constant Ti2,SO of the PI controller expressed as
1
K p2; SO and T i2; SO a2 T 2 (16)
a T 2K2

The parameter a, which constitutes a trade-off between damping of the poles in the closed-loop
transfer function and dynamic responses, can be chosen so as to achieve required damping and de-
sired performance [12,24] as
a 2 1 (17)
However, in literature, the parameter a is constrained between 2 and 4 for optimization according
to SO criterion [23]. The name of the SO criterion comes from the symmetry exhibited by the open-
loop frequency response [25,26].

3. FIELD-ORIENTED CONTROL OF MSC

The main task of MSC is to extract the maximum power from the input source [7]. The eld-oriented
control structure for MSC consists of two control loops(i) inner hysteresis-based current control loop
and (ii) outer PMSG speed control loop. The actual rotor speed r of the PMSG is compared with the

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

reference rotor speed r estimated by the MPPT algorithm, and the speed error is processed through
the outer speed PI controller so as to estimate the reference electromagnetic torque T e as follows
 
  1 t  
T e K ps  r  r r  r d (18)
T is 0
where Kps is the proportional gain and Tis is the integral time constant of the outer speed PI controller.
Transforming Equation (18) in Laplace domain, we have
 
 1 pT is   
T e K ps  r  r (19)
pT is
The reference torque T e so generated is then used to estimate the reference q-axis stator current iqs
using Equation (7). In order to achieve high torque to current ratio as well as to avoid demagnetization
for the surface type PMSG, the reference d-axis stator current ids is typically set to zero [32,33]. Using
inverse Parks transformation of the reference d-q axes stator currents, the reference three-phase stator
currents I sabc are estimated. The actual and reference three-phase stator currents are compared, and the
resulting errors are processed into an inner hysteresis current controller (HCC) to generate switching
signals for the MSC. The switching frequency for the MSC is limited within its specied range by
proper selection of the hysteresis band [41,42].

3.1. Design of speed PI controller


Because hysteresis current control has negligible inertia and delay [30], and the actual current is
limited within the tolerance band of the reference current, the HCC-based current control block can
be viewed simply as a unitary gain in the block diagram of outer speed control loop shown in Figure 2.
The open-loop transfer function of the outer speed control loop is modeled as
   
1 pT is 1 3 P
GOMS p K ps   m P  (20)
pT is 1 pT s 4 p 2J
where Ts is the sample time for speed control loop and J is the turbine-generator mechanical system
inertia. The speed PI controller parameters addressed by SO are calculated as
8J
K ps and T is a2 T s (21)
3 a P2 m T s
On substitution of Equation (21) into Equation (20), GOMS(p) is simplied as
1 p a2 T s
GOMS p     (22)
p2 a3 T 2s p3 a3 T 3s
The closed-loop transfer function of the outer speed control loop is given as
1 p a 2 T s
GCMS p     (23)
1 p a2 T s p2 a3 T 2s p3 a3 T 3s
The response of the outer speed control loop due to a change in the disturbance input is obtained
from the transfer function derived as    
p a3 T 2s P p2 a3 T 3s P
GDMS p      (24)
2J p 2 a2 T s J p2 2 a3 T 2s J p3 2 a3 T 3s J

Figure 2. Outer PMSG speed control loop.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

The respective open-loop Bode plots, locations of closed-loop poles, step and disturbance responses of
PMSG speed control loop with SO tuning of involved PI controller given by Equation (21) for the
chosen values of a = 2, a = 2.4142, a = 3, and a = 4 are shown in Figure 3, and the outcomes are listed
in Table I. It can be observed that the increment in parameter a increases the PM, which, in turn, results
in decreased sensitivity toward changes in the systems parameter and, consequently, the improved
systems robustness. Moreover, more favorable closed-loop pole-zero placements can be obtained
through variation in the parameter a. A lower value of a (< 2.4142) results in poor damping of poles
while a higher value of a (2.4142) leads to improved damping but slower system response. Thus, the
PM should be at least 45 so as to obtain systems dynamic response without oscillations.
It is also analyzed that the SO tuning of PI speed controller gives tracking performance with large
overshoot caused by the forcing element in the numerator of GCMS(p). So, as another choice to decrease
the overshoot and to enhance the performance of the controller, a rst-order pre-lter GFMS(p) on the
reference signal is employed [12].
1
GFMS p (25)
1 p a 2 T s
Thus, the closed-loop transfer function GCMS(p) reduces to
1
GCFMS p     (26)
1p a2 T s p2 a3 T 2s p3 a3 T 3s
It is worth noticeable that the disturbance rejection capability is not affected because the pre-lter
acts only at the reference signal outside of the control loop. Usually, a trade-off between the robustness
and the control performance is taken into consideration while synthesizing the PI controller [21]. Con-
sequently, the optimum selection of the speed PI controller parameters would be a compromise among

Figure 3. (a) Bode plot of open-loop transfer function and (b) root-locus of closed-loop transfer function for
PMSG speed PI controller design addressed bySO with a = 2; SO with a = 2.4142; SO with a = 3; and SO
with a = 4. Performance of outer speed control loop(c) step responses and (d) disturbance rejection capability.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
Table I. Outcomes of outer speed PI controller design.

Copyright 2015 John Wiley & Sons, Ltd.


Step response
Disturbance
Whether Without pre-lter With pre-lter rejection capability
Crossover closed
Tuning GM PM frequency loop Damping Overshoot Settling Overshoot Settling Undershoot Settling
criterion Kps Tis (dB) (deg) (rad/s) stable? ratio (%) time (ms) (%) time (ms) (%) time (ms)
SO (a = 2) 7.12 0.0020 Inf 36.9 1000 Yes 0.5 43.4 8.3 8.15 6.6 0.03 9.4
SO (a = 2.4142) 5.90 0.0029 Inf 45.0 828 Yes 0.707 33.6 7.5 1.40 5.9 0.03 8.7
SO (a = 3) 4.75 0.0045 Inf 53.1 667 Yes 1 24.9 11.8 11.3 0.04 12.7
SO (a = 4) 3.56 0.0080 Inf 61.9 500 Yes 1 17.3 20.5 23.8 0.05 26.5
OPTIMUM DESIGN OF PI CONTROLLERS

DOI: 10.1002/etep
Int. Trans. Electr. Energ. Syst. (2015)
S. M. TRIPATHI ET AL.

the performances as reviewed in Table I. Hence, the selected transfer function of speed PI controller is
given by  
1 0:0029p
Gcs p 5:90 (27)
0:0029p

4. VOLTAGE-ORIENTED CONTROL OF GSC

The main task of GSC is to synchronize the WECS with the grid ensuring high quality of power
delivered to the grid and to control the dc-link voltage [7]. The voltage-oriented control structure for
GSC consists of two cascaded control loops(i) inner current control loop and (ii) outer dc-link
voltage control loop. The actual dc-link voltage VDC is compared with the reference value V DC , and
the voltage error is processed through the outer dc-link voltage PI controller so as to generate the
reference d-axis grid current idg as
 
1 pT iv   
idg K pv  V DC  V DC (28)
pT iv
where Kpv is the proportional gain and Tiv is the integral time constant of the outer dc-link voltage PI
controller. In order to achieve unity-power factor operation of the WECS, the reference q-axis grid cur-
rent iqg is typically set to zero [32,37]. By using Parks transformation of the three-phase grid currents
Igabc, the actual d-q axes grid currents idg and iqg are estimated. The actual and reference d-q axes grid
currents are compared, and the resulting errors are processed into the inner current PI controllers so as
to generate reference d-q axis GSC voltages. For grid synchronization of WECS, a phase-locked-loop
technique [43] is used.

4.1. Design of inner current PI controllers


The grid voltage equations in d-q reference frame, keeping grid voltage vector aligned to the d-axis
[23], are
didg
edg Rg idg Lg  g Lg iqg vdg (29)
dt
diqg
eqg Rg iqg Lg g Lg idg vqg (30)
dt
where g is the angular grid frequency, vdg and vqg are the d-q axes GSC voltages, Rg and Lg are the
coupling resistance and inductance, respectively. To facilitate independent control of d-q axes compo-
nents of grid currents, the cross-couplings due to the coupling inductor are decoupled [12,37] by
dening equivalent control signals as
udg vdg g Lg iqg edg (31)
uqg vqg  g Lg idg eqg (32)
The signals udg and uqg are derived from the inner current control loop as
 
 1 pT ic 
udg K pc  idg  idg (33)
pT ic
 
1 pT ic 
uqg K pc  iqg  iqg (34)
pT ic
where Kpc is the proportional gain and Tic is the integral time constant of the inner current PI
controllers.
The decoupled d-q axes reference GSC voltages are then expressed as follows
vdg udg g Lg iqg edg (35)

vqg uqg  g Lg idg eqg (36)

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

By using inverse Parks transformation, Equations (35), (36) are transformed into the reference a-b-
c voltages and are then applied to the PWM controller to generate switching signals for the GSC [32].
By using Equations (29)(32), the process transfer function of d-q axes current control loops is ob-
tained as
 
I dg p I qg p K gc
(37)
U dg p U qg p 1 pT gc
where Kgc = 1/Rg and Tgc = Lg/Rg.
The open-loop transfer function of the inner current control loops as shown in Figure 4 is modeled
as
   
1 pT ic 1 K gc
GOGC p K pc   (38)
pT ic 1 pT sc 1 pT gc
where Tsc = Tc + 0.5 Tc = 1.5 Tc is the sum of minor time constants and Tc is the sample time for inner
current control loop.
The current PI controller parameters addressed by MO are calculated as
T gc
K pc; MO and T ic; MO T gc (39)
2 T sc K gc
On substitution of Equation (39) into Equation (38), GOGC(p) with MO tuning of PI controller is
simplied as
1
OGC p
GMO   (40)
p 2 T sc p2 2 T 2sc
The closed-loop transfer function of inner current control loop with MO tuning of PI controller is
then expressed as
1
CGC p
GMO   (41)
1 p 2 T sc p2 2 T 2sc
With MO tuning of PI controller, the response of the inner current control loop due to a change in
the disturbance input is obtained from the transfer function derived as
   
p 2 T sc K gc p2 2 T 2sc K gc
GDGC p 
MO       (42)
1 p 2 T sc T gc p2 2 T 2sc 2 T sc T gc p3 2 T 2sc T gc
On the other hand, the current PI controller parameters addressed by SO are calculated as
T gc
K pc; SO and T ic; SO a2 T sc (43)
aT sc K gc
On substitution of Equation (43) into Equation (38), GOGC(p) with SO tuning of PI controller is sim-
plied as
1 p a2 T sc

OGC p 2 3 2
GSO    (44)
p a T sc p3 a3 T 3sc
The closed-loop transfer function of inner current control loop with SO tuning of PI controller is
then expressed as
1 p a2 T sc
CGC p
GSO     (45)
1 p a2 T sc p2 a3 T 2sc p3 a3 T 3sc

Figure 4. Inner current control loop.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

With SO tuning of PI controller, the response of inner current control loop due to a change in the
disturbance input is obtained from the transfer function derived as
   
p2 a3 T 2sc K gc p3 a3 T 3sc K gc
DGC p
GSO         
1 p a2 T sc T gc p2 a3 T 2sc a2 T sc T gc p3 a3 T 3sc a3 T 2sc T gc p4 a3 T 3sc T gc
(46)
The respective open-loop Bode plots, locations of closed-loop poles, step and disturbance responses
of the inner current control loop with MO-based and SO-based tuning of involved PI controller param-
eters given by Equations (39) and (43), respectively, are shown in Figure 5, and the outcomes are listed
in Table II. It is observed that the MO tuning of PI controller results in the highest PM (=65.5),
whereas the SO tuning with a = 2 results in the lowest PM (=36.9). However, the increment in param-
eter a increases the PM which, in turn, results in decreased sensitivity toward changes in the systems
parameter and, consequently, the improved systems robustness. On the other hand, a lower value of
a (< 2.4142) results in poor damping of poles while a higher value of a (2.4142) leads to im-
proved damping but slower system response. It is also evident that MO tuning of current PI controller
results in good response with small overshoot to a step-change in reference signal, but the disturbance
rejection capability is the worst. On the other hand, the SO tuning of current PI controller leads to
much better disturbance rejection capability. To compensate for the forcing element in the numerator
CGC p, a rst-order pre-lter GFGC(p) on the reference signal is also employed [12].
of GSO
1
GFGC p (47)
1 p a2 T sc

Figure 5. (a) Bode plot of open-loop transfer function and (b) root-locus of closed-loop transfer function for inner
current PI controller design addressed byMO; SO with a = 2; SO with a = 2.4142; SO with a = 3; and SO with
a = 4. Performance of inner current control loop(c) step responses and (d) disturbance rejection capability.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
Table II. Outcomes of inner current PI controller design.

Copyright 2015 John Wiley & Sons, Ltd.


Step response
Disturbance
Whether Without pre-lter With pre-lter rejection capability
Crossover closed
Tuning GM PM frequency loop Damping Overshoot Settling Overshoot Settling Undershoot Settling
criterion Kpc Tic (dB) (deg) (rad/s) stable? ratio (%) time (ms) (%) time (ms) (%) time (ms)
MO 85.33 0.00690 Inf 65.5 6.07 103 Yes 0.707 4.32 0.63 N/A 0.06 26.9
SO (a = 2) 85.33 0.00030 Inf 36.9 6.67 103 Yes 0.5 43.4 1.24 8.15 0.99 0.05 6.6
SO (a = 2.4142) 70.69 0.00044 Inf 45.0 5.52 103 Yes 0.707 33.6 1.12 1.40 0.88 0.06 9.5
SO (a = 3) 56.89 0.00068 Inf 53.1 4.44 103 Yes 1 24.9 1.78 1.69 0.07 13.0
SO (a = 4) 42.67 0.00120 Inf 61.9 3.33 103 Yes 1 17.3 3.07 3.57 0.09 17.8
OPTIMUM DESIGN OF PI CONTROLLERS

DOI: 10.1002/etep
Int. Trans. Electr. Energ. Syst. (2015)
S. M. TRIPATHI ET AL.

CGC p reduces to
Thus, the closed-loop transfer function GSO
1
GCFGC p     (48)
1p a2 T sc p2 a3 T 2sc p3 a3 T 3sc
The disturbance rejection capability is not affected from the employment of pre-lter on the refer-
ence signal. The highest crossover frequency is approximately 10 times lower than GSC switching fre-
quency, which is vital to avoid switching noise interference. To satisfy the criteria of both robustness
and closed-loop performance in the control design, the optimum selection of the current PI controller
parameters would be a compromise among the performances as reviewed in Table II. Hence, the se-
lected transfer function of current PI controller is given by
 
1 0:00044p
Gcc p 70:69 (49)
0:00044p

4.2. Design of the outer DC-link voltage PI controller


Because Tsc is very small, hence, for the analysis of the outer dc-link voltage control loop, T 2sc and T 3sc
can be neglected so as to approximate the inner current closed-loop transfer functions expressed as
Equations (41) and (48) by an equivalent rst-order transfer functions as
1
CGC p
GMO (50)
1 p 2 T sc
1
GCFGC p (51)
1 p a2 T sc
Combining Equations (50) and (51) and representing the approximated inner current closed-loop
transfer function as GGCL(p), we obtain
1
GGCL p (52)
1 pT GCL
where TGCL = 2 Tsc for current PI controller addressed by MO criterion or TGCL = a2Tsc for current PI
controllers addressed by SO criterion.
Thus, the open-loop transfer function GOGV(p) of the dc-link voltage control loop as shown in
Figure 6 is modeled as
    
1 pT iv 1 3 edg 1
GOGV p K pv    (53)
pT iv 1 pT sv 2 V DC pC
where Tsv = Tv + TGCL is the sum of minor time constants and Tv is the sample time for the outer dc-link
voltage control loop.
The dc-link voltage PI controller parameters addressed by SO are calculated as
2 CV DC
K pv and T iv a2 T sv (54)
3 a edg T sv
On substitution of Equation (54) into Equation (53), GOGV(p) is simplied as
1 p a2 T sv
GOGV p     (55)
p2 a3 T 2sv p3 a3 T 3sv

Figure 6. Outer dc-link voltage control loop

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

The closed-loop transfer function of dc-link voltage control loop is then expressed as
1 p a2 T sv
GCGV p     (56)
1 p a2 T sv p2 a3 T 2sv p3 a3 T 3sv
The response of the outer dc-link voltage control loop due to a change in the disturbance input is
obtained from the transfer function derived as
   
p a3 T 2sv p2 a3 T 3sv
GDGV p      (57)
C p a2 T sv C p2 a3 T 2sv C p3 a3 T 3sv C
The respective open-loop Bode plots, locations of closed-loop poles, and step and disturbance
responses of outer dc-link voltage control loop with SO tuning of involved PI controller parameters
given by Equation (54) for the selected values of a = 2, a = 2.4142, a = 3, and a = 4 are shown in Figure 7,
and the outcomes are listed in Table III. With these selections, an identical trend in the performance
characteristics can easily be seen as was discussed for PMSG speed PI controller design. Further, to
compensate for the forcing element in the numerator of GCGV(p), a rst-order pre-lter GFGV(p) on
the reference signal is also employed [12].
1
GFGV p (58)
1 p a2 T sv
Thus, the closed-loop transfer function GCGV(p) reduces to
1
GCFGV p     (59)
1p a2 T sv p2 a3 T 2sv p3 a3 T 3sv

Figure 7. (a) Bode plot of open-loop transfer function and (b) root-locus of closed-loop transfer function for dc-
link voltage PI controller design addressed bySO with a = 2; SO with a = 2.4142; SO with a = 3; and SO with
a = 4. Performance of outer dc-link voltage control loop(c) step responses and (d) disturbance rejection capability.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
Table III. Outcomes of outer dc-link voltage PI controller design.

Copyright 2015 John Wiley & Sons, Ltd.


Step response
Disturbance rejection
Without pre-lter With Pre-lter capability
Whether
Crossover closed Settling Settling Settling
Tuning GM PM frequency loop Damping Overshoot time Overshoot Time Undershoot time
criterion Kpv Tiv (dB) (deg) (rad/s) stable? ratio (%) (ms) (%) (ms) (%) (ms)
SO (a = 2) 0.84 0.0037 Inf 36.9 534 Yes 0.5 43.4 15.5 8.15 12.4 8.29 17.6
S. M. TRIPATHI ET AL.

SO (a = 2.4142) 0.69 0.0055 Inf 45.0 442 Yes 0.707 33.6 14.0 1.40 11.0 9.74 16.4
SO (a = 3) 0.56 0.0084 Inf 53.1 356 Yes 1 24.9 22.2 21.1 11.8 23.8
SO (a = 4) 0.42 0.0150 Inf 61.9 267 Yes 1 17.3 38.3 44.7 15.4 49.7

DOI: 10.1002/etep
Int. Trans. Electr. Energ. Syst. (2015)
OPTIMUM DESIGN OF PI CONTROLLERS

From design point of view, the optimum selection of the dc-link voltage PI controller parameters
would be a compromise among the performances as reviewed in Table III. Hence, the selected transfer
function of dc-link voltage PI controller is given by
 
1 0:0055p
Gcv p 0:69 (60)
0:0055p

5. PERFORMANCE TEST RESULTS AND DISCUSSION

Having designed the parameters of involved PI controllers, the effectiveness of the same is evaluated
analytically through a detailed simulation model built in SimPowerSystem of MATLAB/Simulink by

Figure 8. Performance of the PMSG-based wind energy conversion system under varying wind condition
(a) rotor speed, (b) PMSG stator voltages, (c) PMSG stator currents, (d) d-q axes PMSG stator currents,
(e) generated active and reactive powers, (f) dc-link voltage, (g) grid voltages, (h) grid currents, (i) d-q axes
grid currents, and (j) grid active and reactive powers.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

Figure 9. Performance under symmetrical and asymmetrical-fault conditions when no specic MSC control
is activated to enhance fault ride-through (FRT) capability of the PMSG-based wind energy conversion sys-
tem (a) rotor speed, (b) PMSG stator voltages, (c) PMSG stator currents, (d) d-q axes PMSG stator currents,
(e) generated active and reactive powers, (f) dc-link voltage, (g) grid voltages, (h) grid currents, (i) d-q axes
grid currents, and (j) grid active and reactive powers.

implementing the optimally designed values of PI controller parameters and analyzing the performance
of the PMSG-based WECS both under varying wind conditions and under grid-fault conditions.
Various performance test results are shown in Figures 810.

5.1. Performance under varying wind conditions


The simulation results are shown in Figure 8. An initial wind speed is considered as 8.5 m/s, which
drives the wind turbine to rotate at a speed of 18.5 rad/s. The average active and reactive powers gen-
erated at wind speed of 8.5 m/s are about 3450 W and 0 VAr, respectively. After 0.75 s, as the wind
speed starts increasing gradually from 8.5 to 11 m/s, the PMSG speed also starts increasing and
smoothly tracks the reference PMSG speed, which is recursively calculated using the MPPT algorithm.
The speed is settled at about 24 rad/s, which corresponds to the MPP at wind speed of 11 m/s. The

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

Figure 10. Performance under symmetrical and asymmetrical-fault conditions when specic MSC control
is activated to store the active power surplus in the turbine-generator mechanical system inertia so as to en-
hance fault ride-through (FRT) capability of the PMSG-based wind energy conversion system (a) rotor
speed, (b) PMSG stator voltages, (c) PMSG stator currents, (d) d-q axes PMSG stator currents, (e) generated
active and reactive powers, (f) dc-link voltage, (g) grid voltages, (h) grid currents, (i) d-q axes grid currents,
and (j) grid active and reactive powers.

corresponding increase in PMSG voltages and currents can be noticed, and the average active power
generated increases to about 7250 W, but the average reactive power generated remains zero. After
1.75 s, the wind speed starts decreasing gradually from 11 to 8.5 m/s. The corresponding decrease in
the PMSG speed, voltage, current, and average active power generated can also easily be observed.
The actual d-q axes PMSG stator currents accurately track the reference values and results in balanced
three-phase stator current, demonstrating good dynamics of the MSC controllers.
On the other hand, irrespective of any change in wind speed, the dc-link voltage is quickly stabilized at
its nominal value of 800 V, which conrms the effective control of the GSC. The grid voltage is almost
constant at 415 V (r.m.s.), and the variation in wind speed directly affects the amount of grid currents
and hence the average active power delivered to the grid by the GSC. It is noticeable that the average ac-
tive power delivered to the grid is very close to the maximum captured turbine power at different wind

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

speeds, but the average reactive power delivered to/absorbed from the grid remains zero. Moreover, the
total harmonic distortions (THDs) of grid voltages and currents are found within IEEE-519 standard limit.

5.2. Performance under symmetrical and asymmetrical grid-fault conditions


The symmetrical and asymmetrical short-circuit faults at the grid side give rise to the symmetrical and
asymmetrical grid-voltage dips, respectively [44]. The WECSs are not only required to supply power
but also be able to survive under short-duration system faults and voltage unbalances by delivering ac-
tive and reactive powers to the grid with a specic prole depending on the depth of the grid-voltage
dip [45,46]. Such ability is often referred to as the FRT capability of WECSs. Under a grid-voltage dip,
both the active and reactive current references for the GSC are given by the FRT requirement
demanded by the utility operator [45]. According to the E.ON Netz fault response code [47], which
is taken as the reference in this study, a 50% reduction in the grid voltage demands the reactive current
equal to 100% of the rated system current. The FRT requirement is achieved by properly altering the
grid active current reference to zero and reactive current reference in consistent with E.ON Netz fault
response code [47].

5.2.1. FRT capability of WECS without specic MSC control action. Figure 9 shows the simulation
results when no specic MSC control action is taken for enhancement in FRT capability of WECS;
i.e., only GSC is controlled to meet the FRT requirement. At time t = 2.53 s, a symmetrical grid-voltage
dip of 50% of its nominal value is considered for 60 ms. During this symmetrical grid-voltage dip, average
values of d and q axes grid currents are observed as 0 A and about 16 A, respectively. The average active
and reactive powers delivered to the grid are noted as 0 W and about 4000 VAr, respectively, which
demonstrates the fulllment of the FRT requirement during symmetrical grid-voltage dip condition.
At time t = 2.84 s, a grid-voltage dip of 50% in phase-a of its nominal value is considered for
60 ms. During this asymmetrical grid-voltage dip, average values of d and q axes grid currents are ob-
served as 0 A and about 5.2 A, respectively. The average active and reactive powers delivered to the
grid are noted as 0 W and about 2100 VAr, respectively, which demonstrates the fulllment of the
FRT requirement during asymmetrical grid-voltage dip condition also. Because of the unbalanced grid
phase voltages, the oscillations of the instantaneous active and reactive powers around their average
values can easily be observed. Because the MSC and GSC are decoupled and no specic MSC control
action is performed for enhancement in FRT capability of WECS, the PMSG speed, voltages, and cur-
rents remain unaffected during both symmetrical and asymmetrical grid-voltage dips. Further, the
PMSG continues to generate the active power resulting in an imbalance between the active power gen-
erated and the same transferred to the grid, which then causes the dc-link voltage to increase uncontrol-
lably. During the recovery of the dc-link voltage after the fault clearance, the GSC transfers more
active power to the grid than the generated. Accordingly, the current idg reaches the limit (16 A) im-
posed by the PI controller during the dc-link voltage recovery.

5.2.2. FRT capability of WECS with specic MSC control action. The performance analysis presented
earlier considers both the symmetrical and asymmetrical grid-voltage dip conditions to test the perfor-
mance of the GSC for FRT when no specic MSC control action is taken for FRT enhancement. It is
worth noticeable that the increased dc-link voltage may lead to system failure or even damage of both
MSC and GSC [46] and, therefore, is intolerable particularly if the duration of grid-voltage dip is long.
For this reason, many methods are proposed in the literature for the FRT enhancement of PMSG-based
WECSs.
A cheap solution with a simple control proposes the dissipation of the active power surplus in a
braking resistor during grid-voltage dips [45,46,48,49]. Another method for FRT enhancement has
been proposed in [45,48,50] using an energy storage system to absorb the active power surplus during
the grid-voltage dip. FRT capability can also be enhanced by storing the active power surplus in the
turbine-generator mechanical system inertia by increasing the rotor speed during the grid-voltage
dip so as to maintain the dc-link voltage constant [46,51].
The analysis on FRT capability of the proposed PMSG-based WECS with specic MSC control ac-
tion for its enhancement is considered in this subsection where the active power surplus is stored in the

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

turbine-generator mechanical system inertia by increasing the rotor speed during the grid-voltage dips
so as to maintain the dc-link voltage constant. For this, the reference electromagnetic torque command
during grid-voltage dip condition is changed to
E ga I ga E gb I gb E gc I gc
T e FRT (61)
r
In this way, whenever a grid-fault condition is detected, the active power generated by the PMSG is
controlled at once to track the grid-side active power prole, thus retaining the dc-link voltage almost
constant. The simulation results are shown in Figure 10 where the differences can primarily be noticed
in MSC control action. No active power is delivered to the grid during the symmetrical and asymmet-
rical grid-voltage dip conditions considered as before achieving the FRT requirement; meanwhile, the
specic MSC control action described earlier forces the generated average active power to zero, which
avoids the uncontrollable increase in dc-link voltage. Thus, the q-axis PMSG current and, accordingly,
the electromagnetic torque reduce to zero. As a result, there subsists a torque mismatch in the turbine-
generator mechanical system, which causes the rotor speed to increase. During recovery of the rotor
speed after the fault clearance, the energy stored in the turbine-generator mechanical system inertia
is delivered to the grid. For that reason, the currents iqs and idg reach the limit (16 A) imposed by
the respective PI controllers during the rotor speed recovery.

6. CONCLUSION

The designs of PI controllers employed in eld-oriented and voltage-oriented control structures for
machine-side and grid-side converters, respectively, of a grid-connected small PMSG-based wind en-
ergy conversion system were presented. Analytical expressions, closed-loop control transfer functions,
and tuning criteria (MO and SO) for the involved PI controllers were also presented. Optimum selec-
tions of controller parameters were based on the preliminary analyses of the system stability and dy-
namic performance, and it was found that the SO tuning with a = 2.4142 of involved PI controllers
in various control loops offered not only the satisfactory dynamic responses without oscillation but
also good robustness margins. The selected PI controller parameters were applied in the simulation
model, and the results were presented for varying wind conditions and discussed as well. Results were
also presented to test the FRT capability of the WECS rst, without specic MSC control action for
FRT enhancement and subsequently, with specic MSC control action in which the active power sur-
plus is stored in the turbine-generator mechanical system inertia by increasing the rotor speed during
the grid-voltage dips so as to retain the dc-link voltage almost constant. With the proposed control
structure, the PMSG-based WECS works satisfactorily both under varying wind conditions and under
symmetrical and asymmetrical grid-fault conditions, conrming the effectiveness of the design of PI
controllers.

7. LIST OF SYMBOLS AND ABBREVIATIONS

7.1. Symbols
Pitch angle
g Grid voltage phase angle
s PMSG rotor position angle
Damping ratio
s Safety factor
Tip-speed ratio
opt Optimal tip-speed-ratio
m PMSG rotor ux
Air density
r Turbine rotational speed / actual rotor speed

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

r Reference rotor speed


s Electrical rotor speed
g Angular grid frequency
a Parameter constituting a tradeoff between damping of the poles in the closed-loop
transfer function and dynamic response with SO tuning of PI controller
A Swept area
C DC-link capacitance
Cp Power coefcient
edg,eqg d - q axes grid voltages
Egabc Three-phase grid voltages
f Grid frequency
fgsw Switching frequency of the GSC
Gcc(p) Transfer function of inner current PI controller
Gcs(p) Transfer function of outer speed PI controller
Gcv(p) Transfer function of outer dc-link voltage PI controller
GP1(p),GP2(p) Benchmark transfer functions
GCFGC(p) Closed-loop transfer function of inner current control loop with pre-lter on the ref-
erence signal
GCFGV(p) Closed-loop transfer function of outer dc-link voltage control loop with pre-lter on
the reference signal
GCFMS(p) Closed-loop transfer function of outer speed control loop with pre-lter on the refer-
ence signal
CGC p
GMO Closed-loop transfer function of inner current control loop with MO tuning of PI
controller
CGC p
GSO Closed-loop transfer function of inner current control loop with SO tuning of PI
controller
GCGV(p) Closed-loop transfer function of outer dc-link voltage control loop
GCMS(p) Closed-loop transfer function of outer speed control loop
DGC p
GMO Disturbance response function of inner current control loop with MO tuning of PI
controller
DGC p
GSO Disturbance response function of inner current control loop with SO tuning of PI
controller
GDGV(p) Disturbance response function of outer dc-link voltage control loop
GDMS(p) Disturbance response function of outer speed control loop
GFGC(p) Transfer function of rst order pre-lter on the reference signal of inner current con-
trol loop
GFGV(p) Transfer function of rst order pre-lter on the reference signal of outer dc-link volt-
age control loop
GFMS(p) Transfer function of rst order pre-lter on the reference signal of outer speed control
loop
GGCL(p) First order approximation of inner current closed-loop transfer function
GOGC(p) Open-loop transfer function of inner current control loop
OGC p
GMO Open-loop transfer function of inner current control loop with MO tuning of PI
controller
OGC p
GSO Open-loop transfer function of inner current control loop with SO tuning of PI
controller
GOGV(p) Open-loop transfer function of outer dc-link voltage control loop
GOMS(p) Open-loop transfer function of outer speed control loop
idg,iqg d - q axes actual grid currents
ids,iqs d - q axes actual stator currents
idg ,iqg d - q axes reference grid currents
ids ,iqs d - q axes reference stator currents
ip Peak-peak permissible ripple current
Ipeak Permissible peak ac line current

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

Igabc Actual three-phase grid currents


Isabc Actual three-phase stator currents
I sabc Reference three-phase stator currents
J Turbine-generator mechanical system inertia
K1,K2 Plant gains
Kp1,MO,Ti1,MO Proportional gain and integral time-constant of the PI controller addressed by MO
tuning criterion with benchmark plant transfer function of the form GP1(p)
Kp1,SO,Ti1,SO Proportional gain and integral time-constant of the PI controller addressed by SO
tuning criterion with benchmark plant transfer function of the form GP1(p)
Kp2,SO,Ti2,SO Proportional gain and integral time-constant of the PI controller addressed by SO
tuning criterion with benchmark plant transfer function of the form GP2(p)
Kpc, MO,Tic, MO Proportional gain and integral time-constant of inner current PI controller parameters
addressed by MO tuning criterion
Kpc, SO,Tic, SO Proportional gain and integral time-constant of inner current PI controller parameters
addressed by SO tuning criterion
Kps,Tis Proportional gain and integral time-constant of outer speed PI controller
Kpv,Tiv Proportional gain and integral time-constant of outer dc-link voltage PI controller
Ld,Lq d - q axes stator inductances
Lg Coupling ac inductance
ma Modulation index
p Complex frequency
Pm Turbine output power
P Number of PMSG poles
R Radius of the turbine blades
Rs PMSG stator resistance
Rg Resistance of the coupling ac inductor
T e Reference electromagnetic torque
T e FRT Reference electromagnetic torque during grid-voltage dip
T1 Dominant time-constant in the benchmark plant transfer function GP1(p)
T2 Parasitic time-constant or minor-time constant in the benchmark plant transfer func-
tion GP1(p) or GP2(p)
Tc Sample-time for inner current control loop
Tv Sample time for outer dc-link voltage control loop
Ts Sample-time for outer speed control loop
Tsc Sum of minor time-constants in open loop transfer function of inner current control
loop
Tsv Sum of minor time-constants in open loop transfer function of outer dc-link voltage
control loop
vdg,vqg d - q axes GSC voltages
vdg ,vqg d - q axes reference GSC voltages
V Wind speed
VDC Actual dc-link voltage
V DC Reference dc-link voltage
V DCripple Permissible ripple dc-link voltage
VLL Line-line r.m.s. voltage on the ac side of the PWM converter

7.2. Abbreviations
r.m.s. Root-mean-square
FOC Field-oriented control
FRT Fault-ride-through
GA Genetic algorithm
GM Gain margin
GSC Grid-side converter

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

HCC Hysteresis current controller


IAE Integral-absolute error
ISE Integral-square error
ITAE Integral-time absolute error
ITSE Integral-time square error
MO Modulus optimum
MPP Maximum power point
MPPT Maximum power point tracking
MSC Machine-side converter
PI Proportional-integral
PID Proportional-integral-derivative
PM Phase margin
PMSG Permanent magnet synchronous generator
PSO Particle swarm optimization
PWM Pulse width modulation
SO Symmetric optimum
THD Total harmonic distortion
VOC Voltage oriented control
VSC Voltage source converter
WECS Wind energy conversion system
Z-N Method Zeigler-Nichols Method

REFERENCES

1. Li S, Haskew TA, Swatloski RP, Gathings W. Optimal and direct-current vector control of direct-driven PMSG
wind turbines, IEEE Transactions on Power Electronics2012; 27(5):23252337.
2. Mirecki A, Roboam X, Richardeau F. Architecture complexity and energy efciency of small wind turbines, IEEE
Transaction on Industrial Electronics 2007; 54(1):660670.
3. Sharma S, Singh B. Control of permanent magnet synchronous generator-based stand-alone wind energy conver-
sion system, IET Power Electronics 2012; 5(8):15191526.
4. Singh M, Chandra A. Application of adaptive network-based fuzzy inference system for sensorless control of
PMSG-based wind turbine with non-linear load compensation capabilities, IEEE Transactions on Power Electron-
ics 2011; 26(1):165175.
5. Akhmatov V. Modelling and ride-through capability of variable speed wind turbines with permanent magnet gen-
erators, Wind Energy 2006; (9):313326.
6. Rajaei AH, Mohamadian M, Dehghan SM; Yazdian A. PMSG-based variable speed wind energy conversion sys-
tem using Vienna rectier, Euro Transactions on Electrical Power 2011; (21):954972.
7. Blaabjerg F, Teodorescu R, Liserre M, Timbus AV. Overview of control and grid synchronization for distributed
power generation systems, IEEE Transactions on Industrial Electronics 2006; 53(5):13981409.
8. Zigmund B, Terlizzi A, Garcia XT, Pavlanin R, Salvatore L. Experimental evaluation of PI tuning techniques for
eld oriented control of permanent magnet synchronous motors, Advances in Electrical and Electronic Engineer-
ing, 114119.
9. Papadopoulos KG, Margaris NI. Extending the symmetrical optimum criterion to the design of PID type-p control
loops, Journal of Process Control 2012; 22:1125.
10. Wai R-J, Lee J-D, Chuang K-L. Real-time PID control strategy for maglev transportation system via particle swarm
optimization, IEEE Trans Industrial Electronics 2011; 55(2):629646.
11. Astrom KJ, Hagglund T. PID Controllers: Theory, Design, and Tuning, 2nd Ed., Research Triangle Park,
NC: Instrum. Soc. Amer., 1995.
12. Chandra Bajracharya, Marta Molinas, Suul JA, Undeland TM. Understanding of Tuning Techniques of
Converter Controllers for VSC-HVDC, Nordic Workshop on Power and Industrial Electronics, June
9-11, 2008.
13. Neath MJ, Swain AK, Madawala UK, Thrimawithana DJ. An optimal PID controller for a bidirectional induc-
tive power transfer system using multiobjective genetic algorithm, IEEE Transactions Power Electronics 2014;
29(3):15231531.
14. ODwyer A. Handbook of PI and PID Controller Tuning Rules, Imperial College Press, UK, 2nd Edition, 2006.
15. Basilio JC, Matos SR. Design of PI and PID controllers with transient performance specication, IEEE Transac-
tions Education 2002; 45(4):364370.
16. Kristiansson B, Lennartson B. Robust tuning of PI and PID controllers using derivative action despite sensor
noise, IEEE Control Systems Magazine 2006; 5569.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
OPTIMUM DESIGN OF PI CONTROLLERS

17. Rovira AA, Murrill PW, Smith CL. Tuning controllers for set-point changes. Instruments and Control Systems
1969; 6769.
18. Tripathi SM. Modern Control Systems: An Introduction, Innity Science Press: LLC, USA, 1st Edition, 2009.
19. Krohling RA, Rey JP. Design of optimal disturbance rejection PID controllers using genetic algorithms, IEEE
Transactions on Evolutionary Computation 2001; 5(1):7882.
20. Gaing Z-L. A particle swarm optimization approach for optimum design of PID controller in AVR system, IEEE
Transactions on Energy Conversion 2004; 19(2); 384391.
21. Li K. PID tuning for optimal closed-loop performance with specied gain and phase margins, IEEE Transactions
on Control Systems Technology 2013; 21(3):10241030.
22. Ho M-T, Lin C-Y. PID controller design for robust performance, IEEE Transactions on Automatic Control 2003;
48(8):14041409.
23. Suul JA, Molinas M, Norum L, Undeland T. Tuning of Control Loops for Grid Connected Voltage Source Con-
verters, 2nd IEEE International Conference on Power and Energy (PECon 08), December 13, 2008, Johor Baharu,
Malaysia, 797802.
24. Bierhoff MH, Fuchs FW. Active damping for three-phase PWM rectiers with high-order line-side lters, IEEE
Transactions on Industrial Electronics 2009; 56(2):371379.
25. Umland JW, Sauddin M. Magnitude and symmetric optimum criterion for the design of linear control systems: what
is it and how does it compare with the others? IEEE Transactions on Industry Applications 1990; 26(3):489497.
26. Kessler C. Das Symmetrische Optimum Regelungstechnik, 1958 (395426).
27. Aydin O, Akdag A, Stefanutti P, Hugo N. Optimum Controller Design for a Multilevel AC-DC Converter System,
in Proceedings of the Twentieth Annual IEEE Applied Power Electronics Conf. and Exposition, APEC 2005, 610
March 2005; 3:16601666.
28. Zigmund B, Terlizzi A, Garcia XT, Pavlanin R, Salvatore L. Experimental evaluation of PI tuning techniques for
eld oriented control of permanent magnet synchronous motors, Advances in Electrical and Electronics Engineer-
ing, 114119.
29. Bharathidasan SG, Kumudinidevi RP; Ravichandran S. Coordinated controller design of PMSG-based wind tur-
bine using response surface methodology and NSGAII, Int. Trans. Electr. Energ. Syst., 2014.
30. Mohod SW, Aware MV. Micro wind power generator with battery energy storage for critical load,IEEE Systems
Journal, 2011.
31. Goel PK, Singh B, Murthy SS, Kishore N. Isolated wind-hydro hybrid system using cage generators and battery
storage, IEEE Transactions on Industrial Electronics. 2011; 58(4):11411153.
32. Singh M, Khadkikar V, Chandra A. Grid synchronization with harmonics and reactive power compensation capa-
bility of a permanent magnet synchronous generator-based variable speed wind energy conversion system, IET
Power Electronics 2011; 4(1):122130.
33. Tiwari AN, Agarwal P, Srivastava SP. Performance investigation of modied hysteresis current controller with the
permanent magnet synchronous motor drive, IET Electric Power Applications 2010; 4(2):101108.
34. Uehara A, Pratap A,Goya T, Senjyu T, Urasaki AN, Funabashi T. A coordinated control method to smooth wind
power uctuations of a PMSG-based WECS, IEEE Transaction on Energy Conversion 2011; 26(2):550558.
35. Zhang H-B, Fletcher J, Greeves N, Finney SJ, Williams BW. One-power-point operation for variable speed
wind/tidal stream turbines with synchronous generators, IET Renewable Power Generation 2011; 5(1):99108.
36. Tiwari AN, Agarwal P, Srivastava SP. Modied hysteresis controlled PWM rectier, IEE Proceedings-Electric
Power Applications. 2003; 150(4): 389396.
37. Teodorescu R, Blaabjerg F. Flexible control of small wind turbines with grid failure detection operating in stand-
alone and grid-connected mode, IEEE Transactions on Power Electronics 2004; 19(5):13231332.
38. Tan K, Islam S.Optimum control strategies in energy conversion of PMSG wind turbine system without mechan-
ical sensors, IEEE Transactions on Energy Conversion. 2004; 19(2):392399.
39. Oldenbourg RC, Sartorius H. A uniform approach to the optimum adjustment of control loops, Transactions of the
ASME 1954:12651279.
40. Rezek AJJ, Coelho CAD, Vicente JME, Cortez JA, Laurentino PR. The modulus optimum (MO) method applied to
voltage regulation systems: modeling, tuning and implementation, Proc. Inter. Conf. on Power System Transients,
Brazil, June 2428, 2001.
41. Loh PC, Bode GH, Tan PC. Modular hysteresis current control of hybrid multilevel inverter, IEE Proc. Electr.
Power Appl. 2005; 152(1):18.
42. Stefanutti W, Mattavelli P. Fully digital hysteresis modulation with switching-time prediction, IEEE Transactions
on Industry Applications 2006; 42(3):763769.
43. Hsich GC, Hung J. Phase-locked loop techniquesa survey, IEEE Transactions on Industrial Electronics 1996;
43:609615.
44. Bollen M, Zhang JLD. Different methods for classication of three-phase unbalanced voltage dips due to faults,
Renewable and Sustainable Energy Reviews 2003; 11(1):5969.
45. Nguyen TH, Lee D-C. Ride-through technique for PMSG wind turbines using energy storage systems, Journal of
Power Electronics 2010; 10(6):733738.
46. Alepuz S, Calle A, B-Monge S, Kouro S, Wu B. Use of stored energy in PMSG rotor inertia for low-voltage ride-
through in back-to-back NPC converter-based wind power systems, IEEE Transactions on Industrial Electronics
2013; 60(5):17871796.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep
S. M. TRIPATHI ET AL.

47. E.ON Netz GmbH, Grid code: high and extra high voltage, Germany, 2006. [Online]. Available: www.nerc.com/
docs/pc/ivgtf/German_EON_Grid_Code.pdf (Accessed on June 17, 2015).
48. Conroy JF, Watson R. Low-voltage ride-through of a full converter wind turbine with permanent magnet genera-
tor, IET Renewable Power Generation 2007; 1(3):182189.
49. Banham-Hall DD, Smith CA, Taylor GA, Irving MR. Meeting modern grid codes with large direct-drive perma-
nent magnet generator-based wind turbineslow-voltage ride-through, Wind Energy 2012; (15):799810.
50. Abbey C, Joos G. Supercapacitor energy storage for wind energy applications, IEEE Transactions on Industry
Applications 2007; 43(3):769776.
51. Sim J-B, Kim K-C, Son R-W, Oh J-K. Ride-through of PMSG wind power system under the distorted and unbal-
anced grid voltage dips, Journal of Electrical Engineering and Technology. 2012; 7(6):898904.

APPENDIX : SPECIFICATIONS FOR PMSG-BASED WECS


Air density: = 1.229 kg/m2; rated wind speed: Vw = 11 m/s; rated turbine power: Pm = 7.68 kW; power
coefcient: Cp max = 0.4412; tip-speed ratio: opt = 5.66; blade radius: R = 2.6 m; stator resistance:
Rs = 1.4 ; d-q axes stator inductances: Ld Lq = 5.8 mH; rotor ux: m = 2.6 Wb; PMSG poles:
P = 12; turbine-generator mechanical system inertia: J = 1 kg - m2; reference dc-link voltage: V DC
800 V ; dc-link capacitance: C = 1000 F; coupling resistance: Rg = 1.85 ; coupling inductance:
Lg = 12.8 mH; grid voltage: Eg = 415 V (r. m. s.); grid frequency: f = 50 Hz.

Copyright 2015 John Wiley & Sons, Ltd. Int. Trans. Electr. Energ. Syst. (2015)
DOI: 10.1002/etep

S-ar putea să vă placă și