Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Comparison and Performance Analysis of Model Predictive Control

Developed by Transfer Function Based Model and State Space Based


Model for Brushless Doubly Fed Induction Generator

Abstract – Model predictive control (MPC) is an important control technique for Brushless doubly-fed induction
generators (BDFIGs) which are commonly used for wind turbines, and its control performance can be affected
by the MPC design. In this study, the performances of the transfer function based model and the state space
based model are compared in MPC design for BDFIG. For this purpose, transfer function based model
predictive control (TFMPC) and state space based model predictive control (SSMPC) were developed for
BDFIG. The vector control of the BDFIG was simulated using the designed MPCs. The simulation results have
shown that TFMPC produces better results than SSMPC. Additionally ,The simulation results clearly show the
effectiveness and good response of TFMPC in both dynamic operation and steady-state operation. TFMPC
reduces power ripple and decreases harmonics, resulting in an improvement in the quality of the electrical
power generated by the BDFIG. The reference value (set point) was brought closer to the set point with TFMPC,
and the duration of the transient condition was also reduced in this system. The study demonstrated that using
the transfer function to calculate the parameters of the MPC can eliminate the drawbacks of other design
models.

Keywords: Brushless Doubly Fed Induction Generator, Model Predictive Control, State Space Model Based
Model Predictive Control (SSMPC), Transfer Function Model Based Model Predictive Control (TFMPC)

1. Introduction nonlinear direct control methods such as direct torque


control (DTC), a direct power control (DPC), and direct
During the previous two decades, power generation voltage control (DVC). The DVC method is reliable and
by wind has seen a lot of development. There are many types inexpensive, there are no external sensors of stator current
of generators that can be used in wind turbines, and each has and no additional parameters, which can improve the control
its own advantages and disadvantages. In the last fifteen performance [7].
years, doubly fed induction generators (DFIG) have been The complexity of control in DVC can be reduced by
used in applications of wind energy, especially variable using the direct power control (DPC), and large tuning
speed [1]-[2]-[3]. Although there are several advantages to overhead can be overcome. Moreover, DPC provides
using DFIG machines, the inclusion of slip rings and brush exceptional dynamic performance. However, the major
gears increases the complexity of wind energy generation. disadvantage of this method is that the ripple of power is
Due to carbon accumulation and brush gear on internal large. Methods of predictive control have shown to
components, these factors necessitate frequent maintenance effectively improve steady-state and dynamic performance.
of the generator [4]. To overcome this problem, we can use a Model predictive control (MPC) is an important control
brushless doubly fed induction generator (BDFIG) which technique for BDFIG. The design of the MPC affects the
provides the advantages of DFIG along with high reliability control performance of the BDFIG. Three approaches-step
of the system and low maintenance costs owing to the lack response based model,a transfer function based model and
of slip rings and brush gears [5]. state space based model-are used for MPC design. The step
The BDFIG is a brand-new type of AC induction response based model is not suitable for BDFIG control.
machine. It resembles the commonly used doubly fed Transfer function based model predictive control (TFMPC)
induction generator (DFIG). The BDFIG consists of two and state space based model predictive control (SSMPC) are
groups of windings in the stator with separate pairings of used for BDFIG’s control.
poles, one of them is the power winding PW and the second Model predictive power control (MPPC) for a brushless
one is the control winding CW, where PW has been doubly fed induction generator (BDFIG) is achieved in [8].
connected to the grid or load and another CW is connected The BDFIG's state-space equations are concluded, where the
to the inverter [6]. Since BDFIG is considered as a new type current of PW, current of CW, and flux of PW have been
of machine, researchers have made great efforts to enhance chosen as state variables. Then, to predict the future behavior
its performance, especially in control. First, they focused on of the power accurately, a complete power model is
scalar control (open-loop & closed-loop), then on vector developed to represent the relationship between the applied
control, and then on nonlinear direct control methods voltage and the power.
including direct voltage control (DVC), direct torque control In [8], the optimization of a predetermined objective
(DTC), and direct power control (DPC), and finally on function which appears the absolute of power error has been
used for achieving MPPC. Its discrete nature is considered in technique for doubly fed induction generators (DFIGs) is
power converters and no additional modulator is needed, presented; the maximum power point tracking (MPPT) can
however, the main disadvantage of this method is that it has be achieved. When compared to traditional switching table-
some complexity due to using discrete nature in power based direct power control (DPC), the MOMPC method
converters. In this study, this problem has been overcome. In achieves faster transient response, better steady-state
[9], a model predictive power control (MPPC) was performance, and lower switching frequency; this method
developed to overcome the main drawback of DTC/DPC, has the disadvantage of not using a laboratory model, while
namely, large torque/power ripple. In each control period, a in [15], novel sensorless model predictive control (MPC)
zero vector is combined with an active vector. In order to method of a wind-driven doubly fed induction generator
achieve model predictive power control (MPPC), where (DFIG) connected to a dc microgrid has been achieved,
BDFIG's state-space equations must be concluded, the PW where the stator is directly connected to the dc microgrid
flux, CW flux, and rotor flux should be nominated as state through a diode rectifier, and the rotor is fed by only a
variables One of the disadvantages of this method is the voltage source converter (VSC). The main advantage of this
estimation of the flux. connection is reduced power converter cost, but in this
The finite-set model predictive power control method method, the non-linear nature of the diode rectifiers have
(FS-MPPC) was developed for the BDFIG in [10]. To avoid been used, so the THD of the stator current is increased.
the tedious process of estimating the control winding flux in A model predictive voltage control (MPVC) method for
the FS -MPPC controller, the flux of power winding (PW), stabilizing the frequency and amplitude of the output
the current of power winding (PW), and the current of voltages in a Brushless Cascade Doubly-Fed Induction
control winding (CW) were selected as state variables. Generator (BCDFIG) under load changing and variable
Moreover, to increase the accuracy of the BDFIG power speed of generator shaft in stand-alone mode has been
control, the rotor circuit's influence was included in the presented in [16]. However, it is a modern method, that is, it
predictive power model and intrinsically considered in the is not used in the case of connecting with the grid. In [17],
FS-MPPC controller (which has long been ignored in the Low-complexity method for model predictive direct
existing controllers for brushless doubly-fed induction power control (MPDPC) has been presented for doubly fed
machines). Moreover, the feasibility and performance of the induction generator (DFIG) under both unbalanced and
FS-MPPC controller are discussed for many cases, but rotor balanced cases.
circuit influence has been considered into the calculations so According to the analysis above, this paper presents a
the complexity of this system has been increased. comparison between TFMPC and SSMPC. To this end, a
In [11], during grid synchronization, for a brushless mathematical model of the BDFIG is created. The
doubly fed induction generator (BDFIG), a model predictive mathematical model of BDFIG can be used to build TFMPC
virtual power control (MPVPC) is developed. To realize the and SSMPC. Then a complete power model is developed to
grid synchronization condition by controlling the virtual compare TFMPC with SSMPC. SSMPC was designed using
power, the characteristics of control for the virtual power are the state space model. The state space model was used to
analyzed. In [11], the flux feedback must be changed obtain matrices that are used to build control loops for
between the grid connection mode and the grid SSMPC. TFMPC was designed using the transfer function
synchronization mode, so that the MPVPC controller is for the BDFIG, which was derived in detail based on the
designed to be suitable for both modes. The MPVPC machine parameters. The transfer function was used to
controller is capable of achieving rapid and smooth grid calculate the parameters of TFMPC by converting the
synchronization as well as outstanding active and reactive derived transfer function of BDFIG to Z coordinate. The
power decoupling control. predicted model coefficients are determined by comparing
In [12], for doubly-fed induction generators (DFIGs),a the Z-transformed system transfer function with the
predictive direct power control is achieved, which predicts generalized state space equation in TFMPC. This step saves
the stator active and reactive power changes of the DFIG time and effort because you do not have to calculate the
within a specific sample period and calculates the needed predicted model coefficients using equations, which is very
rotor voltage to remove power errors at the conclusion of the time-consuming. After calculating the MPC parameters, two
subsequent sampling period. In this method, the required current control loops were designed. Two current control
switching pulses are generated by space vector modulation loops were used to create a control diagram for the BDFIG,
within the fixed sample period, so the frequency of switching which was the final step in this study. Finally, a digital
will be constant. The results have shown that this strategy of simulation was designed using Matlab/Simulink for the two
control increases the effectiveness and robustness of the controllers TFMPC and SSMPC. The article presents
system performance. simulation results and performance comparisons of two
In [13], to achieve stable switching frequency and different model-based MPCs.
enhanced performance of the system, a modulated model
predictive control (MPC) algorithm for BDFIMs is given. 2. Mathematical Model of BDFIG
[13] shows an improvement in power quality as compared to
the traditional finite control set MPC. A variable speed (generator) BDFIG drive system, as
In [14], based on the multi-objective model predictive illustrated in Figure (1), includes two major parts (CW
control (MOMPC) approach, a simple and effective control winding of control and PW winding of power), where the PW
should be connected to the grid at 90% of the machine power, rotor flux;
whereas the CW should be connected through the converter 𝑅𝑠𝑝 , 𝑅𝑠𝑐 , 𝑅𝑟 PW, CW, and rotor resistance (referred to
system at 10% of the machine power which is shown in stator);
Figure (1) [18] -[19]- [20]. 𝐿𝑠𝑝 , 𝐿𝑠𝑐 , 𝐿𝑟 PW, CW, and rotor self-inductance;
A BDFIG can be classified according to rotor types as
𝐿𝑚𝑝 , 𝐿𝑚𝑐 mutual-inductances between PW and rotor, and
nested-loop-cage type [21]- [22], or reluctance type [23]-
[24]. Brushless doubly-fed reluctance machines CW and rotor.
(DFRMs) have slightly different operating principles. BDFIG's mathematical model will be employed in the
Although, A BDFIG with a nested-loop-cage type has more modeling and simulation procedure for BDFIG that we will
complicated and it is not straightforward to manufacture but implement in this study.
it offers more flexible connections in addition to reducing the
space harmonics of the rotor winding .as a result Compared 3. State Space Model Based Model Predictive
with reluctance rotors, nested-loop- cages rotors have the Control (SSMPC) of the BDFIG
advantages of ruggedness, lower impedance, and higher fill
factors. Moreover, it is cheaper to cast a cage rotor using In this section, state space model based model predictive
aluminum or copper[25]. control (SSMPC) for the BDFIG was designed, according to
mathematical model equations for the BDFIG.
To build a state space model based model predictive
control (SSMPC), several steps must be taken which are the
derivation of the state space model, prediction, and then
optimization.

3.1 State Space Models

Fig. 1. Diagram of the BDFIG system [19] Typically, the state space model has been used to
investigate the attributes and stability of systems by
The mathematical model is a set of mathematical connecting the system's inputs and outputs to model
equations that describe the principal operation of the including the constants determined based on the parameters
machine by linking the variables. The mathematical model of systems.
can be used to build the state space model as well as the The SSMPC systems are built using state space model
transfer function model. where state space model can be written as:
To build the mathematical model of BDFIG, we need to
find the mathematical model for each part of the machine 𝒙𝒎 (𝒌 + 𝟏) = 𝑨𝒎 𝒙𝒎 (𝒌) + 𝑩𝒎 𝒖(𝒌) (1)
(power winding PW -control winding CW -rotor) separately
and then combine them into one model.
The resulting dynamic mathematical electrical model of 𝒚(𝒌) = 𝑪𝒎 𝒙𝒎 (𝒌) (2)
BDFIG can be presented as follows:
Where u is the variable of input; y is the output of
𝑞
𝑣𝑠𝑝 process, and 𝑥𝑚 is the variable of state.
We obtain (3), using the difference between two
𝑣𝑠𝑑𝑝
successive samples of (1),
𝑞
𝑣𝑟 𝑥𝑚 (𝑘 + 1) − 𝑥𝑚 (𝑘) = 𝐴𝑚 (𝑥𝑚 (𝑘) − 𝑥𝑚 (𝑘 − 1)) +
=
𝑑
𝑣𝑟 𝐵𝑚 (𝑢(𝑘) − 𝑢(𝑘 − 1)) (3)
𝑞
𝑣𝑠𝑐
𝑑
[ 𝑣𝑠𝑐 ] Equation (3) can be rewritten as:
𝑞
𝑅𝑠𝑝 + 𝑆𝐿𝑠𝑝 𝜔𝑝 𝐿𝑠𝑝 𝑆𝐿𝑚𝑝 𝜔𝑝 𝐿𝑚𝑝 0 0 𝐼𝑠𝑝
−𝜔𝑝 𝐿𝑚𝑝 𝑆𝐿𝑚𝑝 0 0 𝐼𝑠𝑑𝑝 ∆𝑥𝑚 (𝑘 + 1) = 𝐴𝑚 ∆𝑥𝑚 (𝑘) + 𝐵𝑚 ∆𝑢(𝑘) (4)
−𝜔𝑝 𝐿𝑠𝑝 𝑅𝑠𝑝 + 𝑆𝐿𝑠𝑝
𝑞
𝑆𝐿𝑚𝑝 𝜔𝑟 𝐿𝑚𝑝 𝑅𝑟 + 𝑆𝐿𝑟 𝜔𝑟 𝐿𝑟 𝑆𝐿𝑚𝑐 𝜔𝑟 𝐿𝑚𝑐 𝐼𝑟
𝑑
A new vector of state variable has been selected to
−𝜔𝑟 𝐿𝑚𝑝 𝑆𝐿𝑚𝑝 −𝜔𝑟 𝐿𝑟 𝑅𝑟 + 𝑆𝐿𝑟 −𝜔𝑟 𝐿𝑚𝑐 𝑆𝐿𝑚𝑐 𝐼𝑟 connect ∆𝑥𝑚 (𝑘) to output 𝑦(𝑘) as follows:
𝑞
0 0 𝑆𝐿𝑚𝑐 𝜔𝑐 𝐿𝑚𝑐 𝑅𝑠𝑐 + 𝑆𝐿𝑠𝑐 𝜔𝑐 𝐿𝑠𝑐 𝐼𝑠𝑐
[ 0 0 −𝜔𝑐 𝐿𝑚𝑐 𝑆𝐿𝑚𝑐 −𝜔𝑐 𝐿𝑠𝑐 𝑅𝑠𝑐 + 𝑆𝐿𝑠𝑐 ] [ 𝐼𝑠𝑑𝑐 ] 𝑥(𝑘) = [∆𝑥𝑚 (𝑘)𝑇 + 𝑦(𝑘)]𝑇
Where variables are defined as:
𝑣𝑠𝑝 , 𝑣𝑠𝑐 , 𝑣𝑟 vectors of space for PW voltage, CW voltage, Note that
and rotor voltage; 𝑦(𝑘 + 1) − 𝑦(𝑘) = 𝐶𝑚 ∆𝑥𝑚 (𝑘 + 1)
𝐼𝑠𝑝 , 𝐼𝑠𝑐 , 𝐼𝑟 vectors of space for PW current, CW current,
𝑦(𝑘 + 1) − 𝑦(𝑘) = 𝐶𝑚 𝐴𝑚 ∆𝑥𝑚 (𝑘) + 𝐶𝑚 𝐵𝑚 ∆𝑢(𝑘) (5)
and rotor current;
𝜓𝑠𝑝 , 𝜓𝑠𝑐 , 𝜓𝑟 vectors of space for PW flux, CW flux, and
The following state-space model can be obtained by
composing (4) with (5) : 𝐽 = (𝑅𝑠 − 𝑌)𝑇 (𝑅𝑠 − 𝑌) + Δ𝑈 𝑇 𝑅̅ Δ𝑈 (8)
Equation (8) has two terms. The first one is related to the
∆𝑥𝑚 (𝑘 + 1) 𝐴 𝑜𝑚 𝑇 ∆𝑥𝑚 (𝑘) objective of error minimization, while the second one is
[ ]=[ 𝑚 ][ ]+
𝑦(𝑘 + 1) 𝐶𝑚 𝐴𝑚 1 𝑦(𝑘) related to the magnitude of ΔU. Using (7), we can find the
𝐵 optimal ΔU that minimizes J.
[ 𝑚 ] ∆𝑢(𝑘)
𝐶𝑚 𝐵𝑚
𝑇
∆𝑥 (𝑘) 𝐽 = (𝑅𝑠 − 𝐹𝑥(𝑘𝑖 )) (𝑅𝑠 − 𝐹𝑥(𝑘𝑖 )) − 2Δ𝑈 𝑇 Φ𝑇 (𝑅𝑠 −
𝑦(𝑘) = [𝑜𝑚 1] [ 𝑚 ] (6) 𝐹𝑥(𝑘𝑖 )) + Δ𝑈 𝑇 (Φ𝑇 Φ + 𝑅̅ )Δ𝑈 (9)
𝑦(𝑘)
Differentiating (9) with respect to Δ𝑈, the cost function
Where 𝑜𝑚 = [00 … 0] can be obtained as:
(𝐴, 𝐵, 𝐶) are augmented models, note that (𝐴, 𝐵, 𝐶) can
be calculated based on parameters of systems, in SSMPC, 𝜕𝐽
= −2Φ𝑇 (𝑅𝑠 − 𝐹𝑥(𝑘𝑖 )) + (Φ𝑇 Φ + 𝑅̅ )Δ𝑈 (10)
(𝐴, 𝐵, 𝐶) have been calculated in accordance with BDFIG 𝜕Δ𝑈

parameters such as stator resistance, rotor resistance, 𝜕𝐽


inductance of stator leakage, inductance of rotor leakage, For obtaining the cost function, we should write: =
𝜕Δ𝑈
inductance of magnetizing. One of the major advantages of 0 so Δ𝑈 will be:
the TFMPC system is that it eliminates the need to compute
matrices (𝐴, 𝐵, 𝐶) because all we have to do is transform the Δ𝑈 = (Φ𝑇 Φ + 𝑅̅ )−1 Φ𝑇 (𝑅𝑠 − 𝐹𝑥(𝑘𝑖 )) (11)
transfer function of the system to the Z coordinates and then
compare the resulting equations to the state space equations The relationship between optimal solution and set-point is
to determine the (𝐴, 𝐵, 𝐶).
Δ𝑈 = (Φ𝑇 Φ + 𝑅̅ )−1 Φ𝑇 (𝑅̅𝑠 𝑟(𝑘𝑖 ) − 𝐹𝑥(𝑘𝑖 )) (12)
3.2 Prediction of State
Where 𝑟(𝑘𝑖 ) is set-point signal at sample time k.
In order to design the predictive control algorithm, the
second step is to calculate the predicted value. An 4. Transfer Function Model Based Model
optimization window is used to describe this prediction. This Predictive Control (TFMPC) of BDFIG
is described within an optimization window. We have used
𝑘𝑖 as the current time and the length of the optimization In this part, a model predictive control (MPC) algorithm
window is 𝑁𝑝 as the samples number. for the BDFIG was designed using the transfer function.
The control horizon is denoted by 𝑁𝑐 . To calculate the Transfer function model based model predictive control
prediction of state and variables of output, we can use form (TFMPC), like SSMPC, consists of many steps. The main
of a compact matrix where: advantage of TFMPC is that instead of relying on the
complex computations employed in the SSMPC, the transfer
𝑌 = 𝐹𝑥(𝑘𝑖 ) + ΦΔ𝑈 (7) function of the BDFIG is derived and converted into Z
coordinates, then the (𝐴, 𝐵, 𝐶) matrices are calculated by
Note that the dimension of 𝑌 and Δ𝑈 are 𝑁𝑝 , and 𝑁𝑐 comparison between the calculated equations in Z
respectively. coordinates and state space model. First, for the sake of
simplicity, the TFMPC was completely developed for the
Where single-input single-output (SISO) system. Then, the BDFIG
𝐶𝐴 equations were also applied to the predictive control
𝐶𝐴2 algorithm. Finally, two closed loop currents were built: the
𝐹 = 𝐶𝐴3 first one for 𝐼𝑑 while the second one for 𝐼𝑞 . A flowchart for
⋮ the predictive control algorithm and a complete control
[𝐶𝐴𝑁𝑝 ] diagram for the machine was constructed.
The transfer function model based model predictive
𝐶𝐵 0 0 … 0
control (TFMPC) is applied to BDFIG as follows:
𝐶𝐴𝐵 𝐶𝐵 0 … 0
Φ= 𝐶𝐴2 𝐵 𝐶𝐴𝐵 𝐶𝐵 … 0 Firstly, the transfer function for BDFIG is derived as
⋮ follows:
1−𝑒 −𝑠𝑇𝑠
[𝐶𝐴𝑁𝑝 −1 𝐵 𝐶𝐴𝑁𝑝 −2 𝐵 𝐶𝐴𝑁𝑝 −3 𝐵 … 𝐶𝐴𝑁𝑝 −𝑁𝑐 𝐵] 𝐻(𝑆) = (13)
𝑆
1
3. 3 Optimization 𝐺𝑃(𝑆) = (14)
𝑅𝑟 +𝐿𝑟 𝑠
The last step for designing the SSMPC is an optimization 𝐺(𝑍) = 𝐻(𝑆) 𝐺𝑃(𝑆) (15)
which can be calculated through the error function between Substituting (13) and (14) into (15), a transfer function is
reference point and predicted output, it must be as small as obtained as follows:
possible in order to bring the predicted value as near to the 1−𝑒 −𝑠𝑇𝑠 1
set-point as possible. So, the cost function 𝐽 can be defined 𝐺(𝑍) = 𝐻(𝑆) 𝐺𝑃(𝑆) = (16)
𝑆 𝑅𝑟 +𝐿𝑟 𝑠
as:
Where be concluded as:
𝑅𝑟 : Rotor resistance (referred to stator) 𝑥(𝑘 + 1) = 𝐴𝑥(𝑘) − 𝐵𝑔𝑚𝑝𝑐 𝑥(𝑘) + 𝐵𝑔𝑦 𝑟(𝑘) (23)
𝐿𝑟 :Rotor leakage inductance (referred to stator)
𝑥(𝑘 + 1) = (𝐴 − 𝐵𝑔𝑚𝑝𝑐 )𝑥(𝑘) + 𝐵𝑔𝑦 𝑟(𝑘) (24)
By transferring (15) to Z, equation (16) can be obtained
as: Thus, by using the characteristics of the closed loop
equation, eigenvalues of the closed-loop can be calculated.
1 𝑑𝑒𝑡[𝜆𝐼 − (𝐴 − 𝐵𝑔𝑚𝑝𝑐 )] = 0 (25)
𝑍[𝐻(𝑆) 𝐺𝑃(𝑆) ] = (1 − 𝑧 −1 ) 𝑍 [(𝑅 ] (17)
𝑟 +𝐿𝑟 𝑠)𝑆 Note that vector of state variables:
𝑥(𝑘) = [∆𝑥𝑚 (𝑘)𝑇 + 𝑦(𝑘)]𝑇 , 𝑔𝑚𝑝𝑐 = [𝑔𝑥 𝑔𝑦 ]
By simplifying (17), a transfer function can be written as:
0.02924 5. Simulation Model
𝐺(𝑍) = (18)
𝑍−0.9999
The derived transfer function of BDFIG should be In the last step, the model predictive control is added to
converted into state space equations so (18) must be written the complete control scheme of the BDFIG to compare and
as: analyze the performance of SSMPC and TFMPC by
𝑌(𝑍) 0.02924
= 𝑍−0.9999 (19) modeling and simulation with the Matlab program as we can
𝑈(𝑍)
see in figure (2).
𝑌(𝑘 + 1) − 0.9999. 𝑌(𝑘) = 0.02924. 𝑈(𝑘) This scheme consists of three main control loops: one for
𝑌(𝑘 + 1) = 𝑋(𝑘 + 1) speed and two for current one of them for 𝐼𝑞 , while the
𝑌(𝑘 + 1) = 0.02924. 𝑈(𝑘) + 0.9999. 𝑋(𝑘) another one for 𝐼𝑑 , employing compensation of the cross
𝑌(𝑘) = 𝑋(𝑘) terms has been applied by adding cross term to each loop, in
𝑋(𝑘 + 1) = 0.9999. 𝑋(𝑘) + 0.02924. 𝑈(𝑘) (20) addition to neglecting the effect of the voltage source
The necessary parameters were determined by comparing converter and the possible delays in computation or
(20) with (1), and (2). measurements, as we can see in figure 2, the current control
Where loops have been built based on MPC controller, while the
𝐴 = 0.9999 , 𝐵 = 0.02924 , 𝐶=1 speed control has been built based on PI controller.
As we said before, the major disadvantage for TFMPC is The angle 𝜃𝑟 must be estimated for reference frame
the calculation of (𝐴, 𝐵, 𝐶). transformation, in addition to transforming rotor voltage and
After determining (𝐴, 𝐵, 𝐶), two current control loops currents into DQ coordinates since the control must be
were formed: the first one for 𝐼𝑑 and the second for 𝐼𝑞 . performed in DQ coordinates.
To get 𝜃𝑠 , firstly, the angle of vector of stator voltage
4.1 Closed-loop Control System space can be obtained, after that, 90 degrees must be
subtracted from it. To suppress small disturbances or
A closed loop control system can be defined as a system with harmonics, a simple phase-locked loop (PLL) should be
utilized. In addition, (PLL) can be used to provide robustness
a feedback loop (or) a control system that uses a feedback
to the estimation. It is worth noting that the turn ratio on the
signal to generate the output. The stability of this system can stator and rotor should be taken into account in the control
be controlled by the feedback system, these systems are stage. In the control block diagram shown in the figure
highly accurate. In this part, two current control loops are above, rotor currents are referenced to the stator side in
built: first one for 𝐼𝑑 , whereas second one for 𝐼𝑞 . current loops, whereas the currents are converted to rotor-
To construct a closed loop control system, the optimal referenced values in the measurement step, and the voltages
parameter vector Δ𝑈 must be written as follows: are converted before pulses are generated for the inverter.
Δ𝑈 = (Φ𝑇 Φ + 𝑅̅ )−1 (Φ𝑇 𝑅̅𝑠 − Φ𝑇 𝐹𝑥(𝑘𝑖 ))
Where all terms are dependent on the system's 6. Simulation Results
parameters.
The incremental control is only the first element of Δ𝑈 at To compare the performance of SSMPC and TFMPC, a
time 𝑘𝑖 , i.e.: simulation was performed under the same conditions in the
Δ𝑢(𝑘𝑖 ) = [1 0 … 0](Φ𝑇 Φ + 𝑅̅ )−1 (Φ𝑇 𝑅̅𝑠 𝑟(𝑘𝑖 ) MATLAB/Simulink platform. In this section, the simulation
results are given and discussed briefly. In the table (1),
− Φ𝑇 𝐹𝑥(𝑘𝑖 ))
parameters of the BDFIG related to the stator PW are listed.
Δ𝑢(𝑘𝑖 ) = 𝑔𝑦 𝑟(𝑘𝑖 ) − 𝑔𝑚𝑝𝑐 𝑥(𝑘𝑖 ) (21) Note that Sampling time has been equaled to 𝑇𝑆 = 5 ∗
Where 10−6 (𝑠𝑒𝑐)
𝑔𝑦 = (Φ𝑇 Φ + 𝑅̅ )−1 Φ𝑇 𝑅̅𝑠 , Figure (3) shows the results of the simulated voltages for
𝑔𝑚𝑝𝑐 = (Φ𝑇 Φ + 𝑅̅ )−1 Φ𝑇 𝐹 SSMPC and TFMPC. It can be seen that there is an
The vector of state feedback control gain is 𝑔𝑚𝑝𝑐 . The asymmetric voltage dip in the third second. The voltage dip
augmented system equation is is terminated in the fourth second. Note that for the SSMPC,
𝑥(𝑘 + 1) = 𝐴𝑥(𝑘) + 𝐵 Δ𝑢(𝑘) (22) the voltage 𝑉𝑆 is not plotted as it is the same in MPC because
Substituting (21) into (22), a closed loop system can it is the voltage grid.
Fig. 2. Block diagram of predictive control system for 𝐼𝑞 , 𝐼𝑑 current control loop for BDFIG

Table 1. BDFIG's Parameters changes is very large. It is also worth noting that the transfer
function model based model predictive control (TFMPC)
Stator resistance 𝑅𝑆 0.46 (Ω)
currents are smaller than the SSMPC currents, from 𝑡 =
Rotor resistance (referred to stator) 𝑅𝑟 0.7 (Ω) 0.1 𝑡𝑜 𝑡 = 2(𝑠) in this part of transient state. It can be seen
Inductance of stator leakage 𝐿𝑠 3.5 (𝑚H) that the current change has decreased, the value of current
Inductance of rotor leakage (referred 𝐿𝑟 3.5 (𝑚H) has come closer to reference value, the transient state lasts
to stator) for 2 seconds, then the system enters the stability case where
Inductance of magnetizing (referred to 𝐿𝑚 50 (𝑚H) the oscillations are damped. In stability case, figure (4)
stator) proves that TFMPC has fewer harmonics than SSMPC, i.e.,
the quality of power for TFMPC is higher than the quality of
Figure (4) shows the results of the simulated 𝐼𝑞 current power for SSMPC. Figure (4) shows that TFMPC is closer
for SSMPC and TFMPC. It can be seen that the oscillation is to the set point than SSMPC.
particularly large during the start of the transient state, from
𝑡 = 0 𝑡𝑜 𝑡 = 0.1(𝑠), in this part the value of the current
Voltages for MPC controller
500
Vs

-500
0 1 2 3 4 5 6 7 8 9 10
Time(s)
500 500

0 0

-500 -500
3.4 3.5 3.6 3.7 7.6 7.7 7.8
Fig. 3. Simulation results for voltage 𝑉𝑆 for model predictive control (MPC)
Fig. 4. Simulation results of 𝐼𝑞 current for state space (SSMPC) & transfer function model predictive control (TFMPC)

Fig.5.Simulation results of 𝐼𝑑 current for state space (SSMPC) & transfer function model predictive control (TFMPC)

Fig.6.Simulation results of torque for state space (SSMPC) & transfer function model predictive control (TFMPC)

Figure (5) shows the results of the simulated 𝐼𝑑 current In the model TFMPC, the ripple current is
for SSMPC and TFMPC. Figure (5) proves that TFMPC has ∆𝐼𝑑 = 420 − 382 = 38 (𝐴),
fewer harmonics than SSMPC, therefore the power quality where in the SSMPC the ripple current is
of TFMPC is higher than the power quality of SSMPC. ∆𝐼𝑑 = 475 − 372 = 103 (𝐴).
Figure (5) has shown that the ripple in TFMPC is lower than Thus, we can conclude from the preceding considerations
the ripple in the SSMPC. It can be observed that TFMPC is that the power losses in TFMPC are lower than the power
closer to the set point than SSMPC. losses in SSMPC. Subsequently, the efficiency is higher in
The ripple ratio (currents changing) in the model TFMPC.
predictive controller designed by the transfer function model Figure (6) shows the simulated torque results for SSMPC
is smaller than the change of currents in the model predictive and TFMPC. It can be seen that the oscillation is particularly
control (MPC) designed by the state space model, as we can large during the start of the transient state, from 𝑡 = 0 𝑡𝑜 𝑡 =
see in the figures above where the reference value for is 0.15(𝑠), in this part the value of the torque changes is very
𝐼𝑑 =400 A. large. It is also worth noting that TFMPC torque changes are
smaller than the SSMPC torque changes, from 𝑡 = less need for filters, and due to reduced losses, the efficiency
0.15 𝑡𝑜 𝑡 = 2(𝑠) in this part of transient state. It can be increases and the costs are reduced.
observed that the changing of torque has decreased, there is The simulation results have proved that TFMPC provides
a transient condition which lasts for 2 seconds, then the a higher response speed for the system and also reduces the
system enters the stability case where the oscillations are transient condition, which reduces the losses and increases
damped. Figure (6) clearly shows that TFMPC has fewer the efficiency.
harmonics than SSMPC, therefore, the power quality of In table (2), the comparison between TFMPC and
TFMPC is higher than the power quality of SSMPC. It has SSMPC has been summarized.
been noted that TFMPC is closer to the set point than Table 2. Comparison Summary
SSMPC.
TFMPC SSMPC
It is known that parameter changes occur while BDFIGs
are running. Which affects the operation of MPC; so we Saves time and effort Utilize complex equations
have studied the effect of parameter changing to the MPC. so it needs more time
First, we increased the value of stator and rotor resistance by
a specified percentage (10 %), and the simulation results The transient state The transient state
have been obtained. Then, we increased the value of stator completed within 2 completed within 3
and rotor inductance by a specified percentage (10 %), and seconds seconds
the simulation results have been obtained. Finally, we Provides higher response It has normal response
compared the simulation results we obtained with the speed speed
original simulation results to note the difference in MPC
running. It was observed that there was no discernible effect Reduces the transient The transient state
of changing parameters on MPC. Thus, the figures for condition, transient state completed within 3
parameter variations are not given in this paper. completed within 2 seconds
seconds
7. Conclusion The ripple ratio Ripple ratio (currents
(currents changing) is changing) is big
In this research, a comparison and analysis between small which reduces the
TFMPC and state SSMPC are presented. In addition, the losses and increases the
advantages and disadvantages of TFMPC and SSMPC have efficiency
been determined to improve BDFIG's performance. In this Through this study, we can be concluded that when the
study, the advantages of TFMPC have been discussed briefly. process model is not obtained, the traditional controllers like
Simulations have been done with MATLAB/Simulink to PIDS should be implemented. Having a process model is
analyze the steady-state behaviors and transient behaviors of easy to tune, thus, MPC can be implemented. Additionally,
the model predictive control designed based on the transfer TFMPC is suitable for the most sensitive and important
function model (TFMPC) and state space model (SSMPC). applications, while SSMPC can be used for other
The results have shown that TFMPC can achieve good applications. For future studies, this work will be followed
steady-state features, such as: by an experimental validation using dSPACE DS1103
TFMPC saves time and effort since it reduces the or DSP TMS320F28335 processors.
mathematical effort required to get the (𝐴, 𝐵, 𝐶) matrices; in
addition to that the transfer function can be simply concluded References
for the system and converted it to the Z coordinate, then
compared to the state space model to get (𝐴, 𝐵, 𝐶) matrices [1] Prangya Parimita Pradhan and Bidyadhar Subudhi,
which can be used to build control loops.
“Real-time Active and Reactive Power Control of a
TFMPC reaches a steady state faster than SSMPC as
shown in the figures above, where the settling phase for the Doubly-Fed Induction Generator based Wind Energy
TFMPC is completed within 2 seconds, while the settling Conversion System,” in Proceedings of IEEE Region
phase for SSMPC is completed within approx. 3 seconds 10 Conference (TENCON), Malaysia, November 2017.
(note that the transient can be changed depending on the type [2] Mohammadreza Toulabi, Shahab Bahrami and Ali
of parameter 𝐼𝑞 or 𝐼𝑑 or torque). Note that the oscillation is Mohammad Ranjbar, “An Input-to-State Stability
particularly large during the start of the transient state, in this Approach to Inertial Frequency Response Analysis of
part the value of the torque and currents change is very large.
Doubly-Fed Induction Generator-Based Wind
It is also worth noting that TFMPC torque and currents
changes are smaller than SSMPC torque and currents Turbines,” IEEE Trans. on Energy Conversions, vol.32,
changes. Moreover, the current in the model predictive no.4, pp.1418-1431, Dec. 2017.
controller designed based on transfer function has fewer [3] S. Muller, M. Deicke and R. W. De Doncker, “Doubly fed
harmonics after reaching the stabilization state. This means induction generator systems for wind turbines,” IEEE
that the quality of power in TFMPC is higher than the quality Industry Applications Magazine, vol.8, no.3, pp.26-33,
of power for SSMPC. In which leads to the conclusion that Jun. 2002.
TFMPC has better performance, less noise, less losses, and
[4] Kostyantyn Protsenko and Dewei Xu, “Modeling and Trans. Distrib., vol.13, Iss.1, pp.21-29, 2018.
Control of Brushless Doubly Fed Induction Generators [15] Sertac Bayhan, Haitham Abu-Rub and Omar Ellabban,
in Wind Energy Applications,” IEEE Trans. on Power “Sensorless Model Predictive Control Scheme of Wind-
Electronics, vol.23, no.3, pp.1191-1197, May. 2008. Driven Doubly Fed Induction Generator in DC
[5] Xingwei Wang, Hua Lin and Zhe Wang, “Transient Microgrid,” IET Renew. Power Gener., vol.10, Iss.4,
Control of the Reactive Current for the Line-Side pp.514-521, 2016.
Converter of the Brushless Doubly-Fed Induction [16] H. Abdolrahimi and D. ArabKhaburi, “A Novel Model
Generator in Stand-Alone Operation,” IEEE Trans. on Predictive Voltage Control of Brushless Cascade
Power Electronics, vol.32, no.10, pp.8193-8203, Oct. Doubly-fed Induction Generator in Stand-Alone Power
2017. Generation System,” International Journal of
[6] Jianping Gao, Wei Xu, Yi Liu and Kailiang Yu, Engineering,B:applications, vol.34, no.5, pp.1239-
“Improved Control Scheme for Unbalanced Standalone 1249, 2021.
BDFIG Using Dead Beat Control Method,” IEEE, [17] Yongchang Zhang, Jian Jiao, Donglin Xu, Dong Jiang,
pp.4505-4510, 2018. Zhankuo Wang and Chaonan Tong, “Model Predictive
[7] Xinchi Wei, Ming Cheng, Wei Wang, Peng Han and Direct Power Control of Doubly Fed Induction
Rensong Luo, “Direct Voltage Control of Dual-Stator Generators Under Balanced and Unbalanced Network
Brushless Doubly Fed Induction Generator for Stand- Conditions,” IEEE Trans. on Industry App., vol.56,
Alone Wind Energy Conversion Systems,” IEEE Trans. no.1, pp.771-786, Jan. 2020.
on Magnetics, vol.52, no.7, pp.8193-8203, July. 2016. [18] L. Xu, B. Guan, and H. Liu, “Design and control of
[8] Xinchi Wei, Ming Cheng, Wei Hua, Jianguo Zhu, and ahigh-efficiency doubly-fed brushless machine for
Haitao Yang, “Model Predictive Power Control of a wind power generator application,” IEEE Energy
Brushless Doubly Fed Twin Stator Induction Convers. Congr. Expo., Atlanta, GA, USA, pp.2409-
Generator,” IEEE, pp.5080-5085, 2017. 2416, Sep. 2012.
[9] Xinchi Wei, Ming Cheng, Jianguo Zhu and Haitao Yang, [19] Tang, J. Yang, G. Zhang, Y. Sun, S. Ademi, F.
“Model Predictive Power Control of Dual-Stator Blaabjerg,and Q. Zhu, “Sensorless control of brushless
Brushless Doubly-Fed Induction Generator with doubly-fed induction machine using a control winding
Reduced Power Ripple,” IEEE PEDS, vol.52, pp.607- current MRAS observer,” IEEE Trans. on Industry
612, Dec. 2017. Electron.. , 2018.
[10] Xinchi Wei, Ming Cheng, Jianguo Zhu, Haitao Yang and [20] Hamed Gorginpour, Hashem Oraee and Richard A.
Rensong Luo, “Finite-Set Model Predictive Power McMahon, “Performance Description of Brushless
Control of Brushless Doubly Fed Twin Stator Induction Doubly-Fed Induction Machine in Its Asynchronous
Generator,” IEEE Trans. on Power Electronics, vol.34, and Variable Speed Synchronous Modes,” Journal of
no.3, pp.2300-2311, March. 2019. Electromagnetic Analysis and Applications, no.3,
[11] Xinchi Wei, Ming Cheng, Rensong Luo, Litong Xu and pp.490-511, Oct. 2011.
Jianguo Zhu, “Model predictive virtual power control [21] ]R. A. McMahon, X. Wan , E. Abdi-Jalebi, P. J. Tavner,
of brushless doubly-fed induction generator for fast and P. C. Roberts and M.Jagiela, “The BDFM as a generator
smooth grid synchronization,” IEEE Trans. on Power in wind turbines,” Proc. 12th Int.Power Electron.
Electronics, vol.13, Iss.16, pp.3080-3087, 2019. Motion Control Conf., , pp.1859-1869, Aug. 2006.
[12] Dawei Zhi, Lie Xu and Barry W. Williams, “Model- [22] P. C. Roberts, “A study of brushless doubly-fed
Based Predictive Direct Power Control of Doubly Fed (induction) machines,” Ph.D. dissertation, Emanuel
Induction Generators,” IEEE Trans. on Power College, University of Cambridge., England,2004.
Electronics, vol.25, no.2, pp.341-351, Feb. 2010. [23] S. Ademi and M. G. Jovanovic, “Control of emerging
[13] Xuan Li, Tao Peng, Hanbing Dan, Guanguan Zhang, brushless doubly-fed reluctance wind turbine
Weiyi Tang, Weiyu Jin, Patrick Wheeler and Marco generators,” in Large Scale Renewable Power
Rivera, “A Modulated Model Predictive Control Generation (Ser. Green Energy and Technology), Eds.
Scheme for the Brushless Doubly Fed Induction Berlin, Germany: Springer-Verlag, pp. 395-411, 2014.
Machine,” IEEE Journal of Emerging And Selected [24] S. Ademi and M. G. Jovanovic, “A novel sensorless
Topics In Power Electronics, vol.6, no.4, pp.1681-1691, speed controller design for doubly-fed reluctance wind
Dec. 2018. turbine generators,” Energ. Convers. Manage., vol.120,
[14] Jiefeng Hu, Yong Li and Jianguo Zhu, “Multi-objective pp.229-237, Jul. 2016.
model predictive control of doubly-fed induction [25] Wang, Xuezhou; Liu, Dong; Lahaye, Domenico;
generators for wind energy conversion,” IET Gener. Polinder, Henk; Ferreira, Bram, “Comparison of
Nested-Loop Rotors in Brushless Doubly-Fed
Induction Machines,” IEEE, In 19th International
Conference on Electrical Machines and Systems,
ICEMS, pp. 1-6, 2016.

You might also like