Research objectives and contents • Objective of the thesis Design of a neural controller combined with d-axis stator current control for wind energy conversion system using PMSG to achi
Trang 1UNIVERSITY OF TRANSPORT HO CHI MINH CITY
PhD student NGUYEN NGOC ANH TUAN
SMART CONTROL OF PERMANENT MAGNET SYNCHRONOUS GENERATOR FOR GRID-CONNECTED VARIABLE SPEED WIND
ENERGY SYSTEMS APPLICATION
SUMMARY OF DOCTORAL THESIS
Major: Control and automation engineering
Code: 9520216
Ho Chi Minh City – October 2024
Trang 2The dissertation was completed at Ho Chi Minh City University of Transport
Science Instructor 1: Dr Pham Cong Duy
Science Instructor 2 : Dr Luu Hoang Minh
Trang 31
OVERVIEW
1 The urgency of the dissertation
The demand for electrical energy is expected to increase rapidly due to the rapid growth of the global population and the development of large-scale industry In the field of renewable energy, wind energy is considered a clean energy source and is attracting attention from both the industrial and academic sectors due to its high competitiveness At COP26, Vietnam committed to a program to build a low-carbon economy Therefore, research and development of wind power systems is an extremely necessary and urgent task at present
2 Research objectives and contents
• Objective of the thesis
Design of a neural controller combined with d-axis stator current control for wind energy conversion system using PMSG to achieve maximum power and at the same time reduce system cost
• Contents of the thesis
First, study and learn about wind power generator systems in practice From there, study and compare the advantages and disadvantages of each type Second, build a mathematical model describing the wind turbine system using PMSG and a mathematical model of the PMSG generator Third, study and propose a method to control the d-axis stator current for the machine-side converter in the PMSG used in the wind turbine to maximize efficiency to supply the grid side of the wind power system using PMSG Fourth, study and propose an intelligent control algorithm to improve the quality of the maximum power point tracking controller for the machine side of the wind power system using PMSG Finally, study and propose a controller combining the maximum power point tracking controller using neural networks based on intelligent control algorithms and the d-axis stator current control technique for the generator side of the wind power system using PMSG
3 Research object and method
• The research object is a wind turbine using PMSG with a capacity of 2 MW connected to the local grid
• The thesis was carried out with the following research methods: Firstly, analyze and evaluate the research situation: Study the published research results related to the field of wind turbine system control Next, build a mathematical model
Trang 42 describing the wind turbine system and the mathematical model of the PMSG generator In addition, propose a control method: study models, mathematical tools (d-axis stator current control method, optimal search algorithms, neuron control method) Finally, simulate and evaluate the results: verify simulation and evaluate the effectiveness through comparing optimal control methods The proposed controller solution combines a maximum power point tracking controller using a neural network based on intelligent control algorithms and d-axis stator current control techniques for the generator side of the wind energy system using PMSG
4 Thesis Layout
The thesis is organized including an Overview section; Chapter 1 is an overview
of wind energy; Chapter 2 presents methods for controlling d-axis stator current in wind power systems; Chapter 3 introduces intelligent algorithms for maximum power point tracking controllers in wind power systems using PMSG; Chapter 4 presents the design of maximum power point tracking controllers using RBFN for PMSG in wind power systems; Finally, the conclusion and development direction of the thesis are presented
5 Scientific and practical significance
The study shows that the generator-side controller achieves low cost (economic), good response (technical) In addition, it is the basis for practical application of wind generator system control
CHAPTER 1 OVERVIEW OF WIND ENERGY
1.1 Overview of wind energy
1.1.1 Wind energy development situation in the world
Wind power production in the world has grown rapidly over the past ten years at
a rate of 28% per year, the highest among all existing energy sources The progress and development in technology, capacity, efficiency and reliability of wind power stations have continuously increased, while the cost of wind power has been reduced many times
1.1.2 Wind energy development situation in Vietnam
➢ Wind Power Potential
Some assessment studies indicate that Vietnam has the largest wind power potential compared to other countries in the region for developing large-scale wind
Trang 53 power projects The most promising areas across the country are the coastal areas and the Central and Southern provinces
➢ Strengths and weaknesses in wind power development
Vietnam has many windy areas, especially along the coastline of more than 3,000
km In addition to coastal areas, there are also mountainous, highland and rural areas Places with good wind energy are often far from cities, where electricity is needed Thus, the construction location greatly affects the increase in initial investment capital In addition, wind energy is an intermittent source of energy and cannot be stored at large capacity
Figure 1.1 The structure of the wind turbine [1,2]
1.2 Wind energy conversion system structure
The structure of the wind energy conversion system includes the support tower, turbine blades, steering part, gearbox, generator and wind speed measuring part as illustrated in Figure 1.1
1.3 Generator classification in wind energy system
There are five types of wind turbines that have been developed based on fixed speed or variable speed operation, including 5 types Type 1 is based on fixed speed SCIG Type 2 is based on WRIG which operates at semi-variable speed Type 3 is based on DFIG which operates at semi-variable speed Type 4 is based on
Trang 64 WRSG/SCIG/PMSG which operates at semi-variable speed Type 5 is based on WRSG/SCIG/PMSG which operates at variable speed
1.4 Wind turbine model
The formula for calculating the power and torque of a wind turbine is shown in Figure 1.2 Where p is the air density (kg/m3); A is the area of the wind turbine blade (m2) and is calculated as follows: A= πR2; R is the blade length; vw is the wind speed (m/s); Cp is the turbine power coefficient which is a function of the speed ratio α and the blade pitch angle β c1= 0.5176, c2=116, c3= 0.4, c4= 5, c5= 21, c6= 0.0068
1.5 Permanent magnet synchronous generator model
The formula of PMSG in dq coordinate system is shown as Figure 1.2 In which
uds and uqs are the d-axis and q-axis stator terminal voltages; ids and iqs are the d-axis and q-axis stator currents; Rs is the stator winding resistance; ωr = pωt is the rotor
angular velocity; p is the number of pole pairs; λr is the amplitude of the magnetic flux; λds and λqs are the d-axis and q-axis magnetic fluxes; Te is the torque; TL is the load torque; J is the moment of inertia; B is the friction coefficient
Figure 1.2 Wind turbine model andPermanent magnet synchronous generator
1 ( , ) 2
32
Trang 75
CHAPTER 2 d-AXIS STATOR CURRENT CONTROL METHODS IN
WIND POWER SYSTEM 2.1 d-axis stator current control methods
Three d-axis stator current control methods for machine side converter in permanent magnet synchronous generator based on wind energy systems include zero d-axis stator current control (ZDC), unity power factor control (UPF), and constant stator flux control (CSFL), which are shown in Figure 2.1
2.2 Zero d-axis stator current control
In order to implement this control method (i ds = 0), the d-axis component of the stator current i ds is set to zeroand is illustrated in Figure 2.2
Formula for calculating current, voltage, power factor and torque of generator [15,16]
s ds qs qs
i = + i ji = ji khi ids = 0 (2.1)
Where i s the stator current space vector and is represents the magnitude of the stator current, which is also the peak value of the three-phase current in the static reference
frame ρ are the pole pairs and λr is the rotor flux generated by the permanent magnet
in the PMSG, θv and θi, are the angles of the stator voltage and current vectors, the stator resistance Rs, and the generator rotor speed ɷr
2.3 Unity power factor control
Under this control method, the stator current vector and stator voltage vector are
in the same direction Hence, power factor angle becomes zero and is illustrated in Figure 2.3.According to [15,16]
-Speed controller
i*ds =0
i*ds # 0 dq
abc
a,b,c phase stator voltage controllers
Trang 86 UPF condition is
2
r qs
2.4 Constant stator flux-linkage control
The constant stator flux-linkage is used to overcome the problem of increasing stator flux linkage as the torque reference value is increased, which can lead to saturation of the stator yoke and is illustrated in Figure 2.4
Table 2.1 Summary of the relationship between d-axis and q-axis stator currents
in three control methods
Control methods Relationship between d and q axis stator currents
Trang 97 and q axes Therefore, the is current of the CSFL method has a value twice as large
as the is current of the UPF method and this is shown in Figure 2.5
Figure 2.2 Vector diagram of ZDC [16] Figure 2.3 Vector diagram of UPF [16]
Figure 2.4 Vector diagram of CSFL
2.5 Analysis and comparison of three flow control methods
To evaluate the performance of d-axis stator current control methods (ZDC, UPF and CSFL) of the machine-side converter is verified through waveforms of active and reactive power, power factor and current at the generator side under the same operating conditions using Matlab The parameters of the turbine and PMSG generator used in the simulation are referred to [1]
In order to evaluate the advantages and drawbacks of the proposed three stator current control methods, several sets of simulations are conducted using Matlab The
Trang 108 simulation results in Matlab show the waveforms of active and reactive power, power factor and the current in generator side The effectiveness of the ZDC control method, UPF control method and CSFL control method of the machine side converter is examined under different operating conditions The parameters of the PMSG and turbine are given in Tables 1 and 2
Table 2.2 Comparative parameter in three control methods
Figure 2.6.a shows the assumed wind speed in the range of 7m/s to 12m/s Figure 2.6.b shows that the UPF control method (idUPF) has the largest amplitude value in the waveforms of the d-axis stator current, while the ZDC control method (iqZDC) has
a value of 0 at all times Figure 2.6.c The q-axis stator current (iqZDC, iqUPF, iqCSFL) in the three control methods are approximately the same value when the wind speed varies Figure 2.6.d The simulation results of the stator flux The stator flux of the ZDC control method (λZDC) has the largest value, the stator flux of the UPF control method (λUPF) has the smallest value with different wind speeds It can be seen that the stator flux of the CSFL control method (λCSFL) does not change
Figure 2.7 shows that the active power is approximately the same value because according to the formula of Figure 1.2, the torque or active power is proportional to the stator current of the q axis In addition, the reactive power in the UPF control method (QUPF) is always 0, the largest value is the reactive power in the ZDC control method (QZDC) Figure 2.8 The apparent power of the ZDC control method (SZDC)
Trang 119 has the largest value, this reason increases the cost of the ZDC control method, while the UPF control method has the smallest oscillation Figure 2.9 The power factor in the UPF control method has the largest value (θUPF) and the power factor in the CSFL control method (θCSFL) has the average value while the power factor in the ZDC control method (θZDC) has the smallest value Figure 2.10 the stator current of the UPF control method (isUPF) has the largest value so the UPF control method will have
to choose a wire with a large cross-section for the machine-side converter and this reason increases the cost and the ZDC control method (isZDC) has the smallest value From the above analysis, we can summarize as shown in Table 2.2
Figure 2.6(a) Simulation results
regarding the time-domain waveforms
of wind speed
Figure 2.7 Simulation results regarding the time-domain waveforms of active- and reactive power in three control methods
Figure 2.6(b) Simulation results
regarding the time-domain waveforms
of d-axis stator current in three control
methods
Figure 2.8 Simulation results regarding the time-domain waveforms of apparent power in three control methods
Trang 1210 Figure 2.6(a) Simulation results
regarding the time-domain waveforms
of q-axis stator current in three control
methods
Figure 2.9 Simulation results regarding the time-domain waveforms of power factor in three control methods
Figure 2.6(d) Simulation results
regarding the time-domain waveforms
of stator flux in three control methods
Figure 2.10 Simulation results regarding the time-domain waveforms with rms stator current in three control methods
The summarized Table 2.2 is shown that
• ZDC control method has poor performance and the reactive power is high therefore the cost is high during wind speed variations
• UPF control method has high performance during wind speed variations, however, reduce the power rating of the proposed configuration This would lead to
a smaller size and hence reduce the cost of the power circuit, which is one of the significant considerations for megawatt-level wind turbine design
• CSFL control method has approximately high performance and the power factor is approximately one during wind speed variations
2.6 Experimental results
The proposed HIL configuration consists of two parts as illustrated in Figure 2.11
The first part is a computer with RT-Lab software, the second part is the
OPAL-RT (OP5707-XG) which uses an Intel processor for real-time computation of the models
Figure 2.12 shows the actual HIL configuration on OPAL-RT OP5707XG using DSOX 2014A oscilloscope to measure signal results
Figure 2.13 shows the configuration of connecting OP5707XG hardware to DSOX 2014A oscilloscope via two input and output cards OP5340 and OP5330 Figure 2.14 shows the measured results from the DSOX 2014A oscilloscope of wind speed signals and d-axis currents (idZDC,idUPF,idCSFL) of the 3 control methods
Trang 13is proposed to achieve good control quality, small apparent power, high economic value because of reduced cost
Results achieved
Trang 1412
Tuan Ngoc Anh Nguyen, Pham Cong Duy, Cong-Thanh Pham, and Nguyen Huu Chan Thanh,
“D-axis stator current control methods applied to PMSG-based wind energy systems: A comparative study” WSEAS transactions on systems and control, vol
14, Art.#31, pp 239-246, 2019
CHAPTER 3 INTELLIGENT ALGORITHM FOR MAXIMUM POWER POINT TRACKING CONTROLLER IN WIND ENERGY SYSTEMS 3.1 Intelligent algorithm for maximum power point tracking controllers
The relationship between turbine power (Pm) and wind speed (vw) when the wind speed changes Under the condition of rapidly changing wind speed, the traditional MPPT method fails to track the MPP and may get stuck in the local maximum [24] Therefore, these problems have been solved by intelligent algorithm-based MPPT algorithms such as GA, PSO, DE and EO [25-27]
3.2 Genetic Algorithm (GA)
In GA, each person is assigned a fitness rating along each route and the best individuality is chosen as a chromosome Natural selection is the process of producing new optimum individuals in GA, and this cycle is accomplished via repeated applications of genetic operators such as selection, crossover, and mutation The GA flowchart is presented as Figure 3.1
Evaluate the target function and update the
new equilibrium state
Generate new population
50 Maximum number of
generations, itermax
50 Crossover probability 0.75 Mutation probability 0.1 Length of chromosome 2
Trang 1513
3.3 Particle Swarm Optimization (PSO)
The PSO algorithm is like a genetic algorithm in that the first phase is initialization, in which a population of random solutions is used to produce the first swarm of particles, each particle has two primary properties: location and velocity The PSO flowchart is presented as Figure 3.2
Evaluate the target function and
Update and calculate new variable
w ri = w rmin + rand(w rmax - w rmin )
v i = w min + rand(w max - w min )
Stop Check stop condition
iter <=itermax
iter = iter +1
Figure 3.2 Flowchart of PSO
Bảng 3.2 PSO parameters used in simulation
Particle number of a generation, Npop
50 Maximum number of
generations, itermax
50
Acceleration Coefficient (c1, c2) c1=c2=2 Independent Random
Sequence (r1, r2) rand (0,1)
Position of initial particles rand (0,1)
Velocity of initial particles rand (0,1)
3.4 Equilibrium optimizer algorithm (EO)
The EO is a new physical-based optimization algorithm that is proposed in 2019
by Faramarzi et al The EO algorithm can mutate the random solving a problem via exploration and exploitation A particle with its concentration is operated similar a solution and position in the PSO algorithm and update its concentration with particular terms It defined as best-so-far solution, called the equilibrium candidate and the other is equilibrium state, which encourages particle to global search the domain.Công thức vi phân bậc nhất biểu thị công thức cân bằng khối lượng chung
equ
• Initialization and function evaluation
min ( max min); 1, 2 ,
initial
Trang 1614 where Ciinitial is the initial concentration vector of ith particles, Cmin and Cmax are the minimum and maximum value of dimensions, randi is a random number in [0, 1], and n is the total number of particles
• Equilibrium pool and candidates (C eq )
where λ is assumed to be a random vector in the interval of [0, 1] and t is a
function of iteration that decreases with the number of iterations
2
_1
_
iter Max iter Iter
where Iter and Max_iter present the current and the maximum number of
iterations, respectively, and a2 is a constant value used to manage exploitation ability The calculation of t0 is as follows
The generation rate (G) enables the EO algorithm to provide accurate solutions
by improving the exploitation phase
Trang 17Evaluate the target function and
update the new equilibrium state
Update and calculate new
variable wr-new
No
Yes Stop
Input Vw = rand([3,12],120)
Start Initialization of EO s parameters
a1, a2, Cp,
Initialization of EO s initial
population wri = wrmin + rand(wrmax - wrmin)
Check for stop
50 Maximum number of
generations, itermax
50 Constant used to manage
3.5 Differential Evolution (DE)
In DE algorithms for optimizing functions in an N-dimensional continuous area Every type in the population is an N-dimensional vector that represents the problem solving DE is based on taking the differentiation vector between two kinds and a scaled version of the distinguishing vector was added to a third person to produce another applicant arrangement
Flowchart of DE
Trang 1816
Initialization of an initial equilibrium state
w r1, w r2, w r3, FX 1, FX 2, FX 3,
Evaluate the target function and update the
new equilibrium state
Update and calculate new variable (w r-new )
No
Yes Stop
Input V w = rand([3,12],120)
Start
Initialization of DE s parameters
Initialization of DE s initial population
w ri = w rmin + rand(w rmax - w rmin )
Check for limit of variable (w r )
Pop = w rmin + rand (w rmax – w rmin)
Check stop condition
Figure 3.4 Flowchart of Differential
50 Maximum number of
generations, itermax
50 Crossover probability 0.2 Lower Bound of
Upper Bound of
1 Determine the parameters of the algorithm
2 Initialize the population
where xn,i and yn,i are the minimum and maximum values of the design variable
i, i = 1,2, and rand (0,1) is a randomly initialized real number in the range [0,1]
3 Mutation
where X1, X2 and X3 are three randomly selected vectors from the set and F is
a geometric progression which is the basic boundary of the DE operation
4 Crossover