Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/254017744
CITATIONS READS
5 7
7 AUTHORS, INCLUDING:
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Marco Rivera
letting you access and read them immediately. Retrieved on: 03 February 2016
Rotor Current Fuzzy Control of a DFIG with an
Indirect Matrix Converter
C. F. Calvillo1, F. Martell1, J. L. Elizondo1, A. Ávila1, M. E. Macías1, M. Rivera2, J. Rodriguez2
1
Tecnológico de Monterrey, Campus Monterrey, Monterrey, Nuevo León, México
2
Universidad Técnica Federico Santa María, Valparaíso, Chile
c.f.calvillo@ieee.org, fmartell@ingmt.com, jl.elizondo.carrales@itesm.mx,
aavila@itesm.mx, mmacias@itesm.mx, marco.rivera@usm.cl, jrp@usm.cl
Abstract—This paper presents a soft-computing technique to The Doubly Fed Induction Generator (DFIG) is a feasible
control the rotor current of a Doubly Fed Induction Generator option for application in Wind Energy Conversion Systems
(DFIG) using an Indirect Matrix Converter (IMC). The tech- (WECS). DFIGs offer a flexible range of operation speed
nique developed uses fuzzy logic to successfully control the and control. The power converter is rated at only 1/3 of the
rotor current, even with abnormal grid conditions, by govern-
generator capacity and the DFIG can deliver up to 1.3 times
ing the IMC rectifier and inverter sides. Although this control
problem has been already analyzed and solved with conven- the nominal power [3]. Moreover, while an AC-AC convert-
tional vector control and with Finite States Model Based Pre- er feeds the rotor circuit and the stator is directly connected
dictive Control (FS-MBPC), the fuzzy logic based controllers to the grid, a DFIG allows supersynchronous and subsyn-
have a very acceptable response. The advantages of the devel- chronous shaft speeds [3], [4].
oped technique are simpler code without the need of a specific Conventionally the power converter is designed consider-
mathematical model and robustness under grid abnormalities. ing a Back-to-Back Voltage Source Converter (BBVSC)
These benefits decrease system costs, increase overall reliability configuration, controlled with Space Vector Modulation
and allow the control scheme to be implemented easily and
entirely in FPGAs instead of DSPs-FPGA control systems.
(SVM), which has an intermediate DC-Link [5]. The indi-
rect matrix converter (IMC) consists of a two-stage convert-
Index Terms—Doubly Fed Induction Generator (DFIG), Indi- er that directly connects the power supply to the load with-
rect Matrix Converter (IMC), Fuzzy Control out a DC-Link or large energy storage elements, making it
an alternative to traditional back-to-back converters [6]. It
NOMENCLATURE
also offers a major advantage when size is a crucial issue.
λs, λr Stator and rotor linkage flux. Years of continuous research have been dedicated to the
Rs, Rr Stator and rotor resistance. development of new IMC topologies (Sparse, Very Sparse,
Ls, Lr Stator and rotor self-inductance. Ultra Sparse) for specific applications, new semiconductors,
Lm Mutual inductance. and different modulation and control strategies for direct
D Euler derivative operand. matrix converters (DMC) and IMCs, as reported in [7]-[10].
p Generator poles number. The IMC’s absence of passive components makes it a relia-
J, F Generator inertia and friction. ble smaller ―pure-silicon‖ converter. Additionally, matrix
Srx, Siy Rectifier and inverter switch state. converters allow bidirectional power flow from and to the
vin, iin IMC input voltage and current. rotor, which is needed for subsynchronous and supersyn-
vdc, idc IMC DC-Link voltage and current. chronous operation speeds [10]. Fig. 1 shows the conven-
Superscripts tional configuration of the IMC to be addressed.
* Reference value. The rotor current control in DFIG-based WECS applica-
Subscripts tions is relevant because its magnitude, frequency, sequence
s, r, m, f Stator, rotor, mechanical, filter. and phase directly control the active and reactive powers
d, q dq reference frame axis. delivered to or absorbed from the grid through the stator.
e, T Electrical, wind turbine. The rotor current must be controlled in real time for chang-
ing wind or grid conditions.
I. INTRODUCTION Nowadays, model predictive control has been successful-
In recent years, wind power generation has experienced ly applied in power converter control as an alternative to
significant growth worldwide. The market value of wind conventional SVM and its associated vector control tech-
power can benefit from an increasing demand for electricity niques, which require the selection of a set of space vectors
and from rising fuel and CO2 prices [1]. Due to the econom- to trigger the power electronics devices in a proper sequence
ic impact of wind power, efficient control techniques are [11]. Therefore, a Finite States Model Based Predictive
becoming a focus of the world’s renewable energy challeng- Control (FS-MBPC) has been introduced for the current
es, making it a promising field of study. With better control- control and instantaneous reactive power minimization in an
lers and the constant development of new technologies, IMC [12]. In particular, the rotor current control of a DFIG,
wind power will maintain its rapidly growing market ratio with a BBVSC and an IMC, can be achieved very precisely
[2]. through FS-MBPC as reported in [9], [13], respectively.
Wind
Turbine
DFIG
As vs
ωT Grid
is
ωm As
br ar Bs
Cs
Bs Cs
cr
Gearbox cr br ar
Inverter DC-Link Rectifier
+P
Sr5 Sr3 Sr1
vr Si5 Si3 Si1 vin vf
vdc
Fig. 1. Wind energy conversion system with DFIG technology controlled by an IMC on the rotor side [9].
switching states for the rectifier and the inverter parts of a ir*
Si Sr
conventional IMC, respectively, as reported in [9].
ir ErrA, B, C
Inverter-side
Delay
Rectifier-side
vinA, vinB, vinC
+
- Fuzzy Fuzzy
III. PROPOSED FUZZY CONTROL 3 3
Controller Sr ’ Controller
3
the next switching state for the inverter. None of these con- c)
trollers need any extra modulation steps. Fig.4. Membership functions for the inputs (a), (b) and the output (c) of the
rectifier-side fuzzy controller.
TABLE I
BN MN SN Z SP MP BP
RULE BASE OF THE RECTIFIER-SIDE FUZZY CONTROL 1
Membership
Degree of
PrevState = P1
S7 S5 S1 S1 S1 S1 S1 S1 P9 a)
BN S1 S2 S3 S4 S5 S6 S7
1
S5 S5 S5 S5 S1 S5 S1 S1
Membership
Degree of
S4 S6 S2 S8 S2 S1 S2 S1
SN
S6 S6 S5 S5 S1 S5 S1 S1
vinB 0 1 2 3 4 5 6 7
S4 S3 S4 S3 S2 S2 S2 S2 Inverter’s Next State
SP b)
S6 S6 S6 S6 S9 S5 S1 S1
Fig. 5. Membership functions for the inputs (a) and the output (b) of the
S3 S3 S4 S3 S4 S4 S2 S2 inverter-side fuzzy controller IFC.
BP
S3 S3 S3 S3 S4 S6 S2 S7
TABLE II
RULE BASE OF THE INVERTER-SIDE FUZZY CONTROL
ErrC
BN MN
Each switching state (S1-S9) delivers a two voltage dif- SN Z SP ErrA
ference from the three phases. Therefore, the fuzzy ―If- MP BP BN MN SN Z SP MP BP
Then‖ rules are designed in accordance with the premise of BN S6 S5 S5 S5 S5 S5 S1
always having the maximum positive voltage in the ficti- S4 S6 S4 S5 S4 S5 S4 S5 S4 S1
tious DC-Link of the IMC while avoiding excessive switch- MN S6 S6 S5 S5 S5 S1 S1
ing. The inference method used is Mamdani’s ―Max-Min‖ S6 S3 S5 S3 S5 S3 S5 S3 S1 S3
approach and, due to the discrete nature of the output, the S4 S6 S4 S4 S4 S4 S4 S4 S4 S1
Mean of Maximum (MoM) method is used for defuzzifica- SN S6 S6 S6 S5 S1 S1 S1
S6 S3 S3 S3 S3 S3 S3 S3 S1 S3
tion. The controller implies 245 rules (Table I), delivering a
S4 S6 S4 S4 S4 S4 S4 S4 S4 S1
robust control for the converter rectifier-side.
ErrB Z S6 S6 S6 S4 S7 S3 S1 S1 S1
B. Inverter-Side Fuzzy Control S6 S3 S3 S3 S3 S3 S3 S3 S1 S3
S4 S6 S4 S4 S4 S4 S4 S4 S4 S1
The IFC complements the control system, which uses the
SP S6 S6 S6 S2 S1 S1 S1
three current errors as inputs: ErrA, ErrB and ErrC. These
S6 S3 S3 S3 S3 S3 S3 S3 S1 S3
errors are calculated by the three-phase current reference S4 S6 S4 S2 S4 S2 S4 S2 S4 S1
minus the actual rotor current, as in Fig. 3. The IFC takes MP S6 S6 S2 S2 S2 S1 S1
these inputs and delivers one of the valid switching states to S6 S3 S2 S3 S2 S3 S2 S3 S1 S3
the IMC inverter side. BP S6 S2 S2 S2 S2 S2 S1
Because the error of the rotor current can be of any
magnitude, the universe of discourse for the membership
functions goes from ± ∞, but the MFs are more precise proposed controller in order to validate its performance. The
nearing the zero, as apparent in Fig. 5a. For each input, simulations were developed in MATLAB/Simulink with the
seven MFs were defined with the following terms: BN, MN following scenarios: constant shaft speed, constant shaft
(medium negative), SN, Z (zero), SP, MP (medium positive) speed with abnormal grid conditions and finally, constant
and BP. In the outputs case, there are 8 valid states for the torque with variable speed. All the simulation variants con-
inverter: six ―active‖ and two ―zero‖ states. The two ―zero‖ sidered zero initial conditions.
switching states are considered as one state, leaving the It is important to note that these simulations were carried
output defined in seven MFs (S1 - S7) as can be seen in Fig. out to fully test the proposed control scheme, involving
5b. Each switching state, when applied, reduces the positive arbitrary rotor current references, nominal and extreme shaft
or negative error of the rotor current in one phase. velocities and torques, unbalanced and phase shifted input
Based on the knowledge of the system, 139 control rules voltage, ideal reactive elements and parameters as in [9].
were defined (Table II) in order to first reduce the greatest A. Step references at constant speed
current error of the three inputs. When two or three of the
Constant supersynchronous and subsynchronous speeds
errors have the same magnitude, the controller gives priority
were first considered for testing. To start, a reference of 1/9
to one of the phases, minimizing extra stress delivery to any
particular IGBT. Similar to the RFC, the Mamdani’s ―Max- of the nominal stator current at 18 Hz from t = 0 s to t = 0.25
Min‖ inference machine and the defuzzification method s was tested, and after that moment a step was imposed,
MoM is used in the IFC. changing the reference to 1/3 of the nominal value.
The transient response at the beginning of the simulations
IV. SIMULATION RESULTS derives from the initial zero conditions. Fig. 6a shows the
behavior of the system with supersynchronous shaft speed
The scope of this research work is the simulation of the
Fig. 6. Rotor current fuzzy control with constant speed = 2340 rpm.
Fig. 8. Rotor current fuzzy control with constant speed = 2340 rpm and
unbalanced grid conditions.
Fig. 7. Rotor current fuzzy control with constant speed = 1260 rpm.