Study of identified and control permanent magnet synchronous motor using the algorithm neural network (nghiên cứu nhận dạng và điều khiển động cơ đồng bộ nam châm vĩnh cửu sử

https //iaeme com/Home/journal/IJPTM 9 editor@iaeme com International Journal of Production Technology and Management (IJPTM) Volume 13, Issue 1, January December, 2022, pp 9–16, Article ID IJPTM 13 0[.]

Available online at https://iaeme.com/Home/issue/IJPTM?Volume=13&Issue=1

ISSN Print: 0976- 6383 and ISSN Online: 0976 – 6391

DOI: https://doi.org/10.17605/OSF.IO/32RG6

© IAEME Publication

STUDY OF IDENTIFIED AND CONTROL

PERMANENT MAGNET SYNCHRONOUS MOTOR USING THE ALGORITHM NEURAL

NETWORK

Nam Nguyen Duy

Dong Nai Technology University, Vietnam

ABSTRACT

This paper proposes a method to control speed of permanent magnet synchronous motor (PMSM), using the PI control combines intelligent neural network control, radial basis function neural network (RBF NN) Firstly, the model of RBF NN, consisting of

an input layer, a hidden layer of nonlinear processing neurons with Gaussian functions and an output layer is presented in the paper In particular, a monitoring mechanism based on the Stochastic gradient descent (SGD) method is used to update the parameters of the RBF NN to minimize the error to minimum the error value Then, the high-speed integrated circuit hardware description language (VHDL) is used to describe the behavior of the whole RBF NN and related learning algorithms Simulation model is implemented in Matlab Simulink and ModelSim Finally, the RBF neural network is used for identification and control in linear/ nonlinear systems and in permanent magnet synchronous motor systems (PMSM), to validate the efficiency and effectiveness

Key words: Radial Basis Function Neural Network (RBF NN); PI Controller; VHDL;

Simulink simulation; ModelSim simulation; Permanent magnet synchronous motor (PMSM) drive

Cite this Article: Nam Nguyen Duy, Study of Identified and Control Permanent

Magnet Synchronous Motor Using the Algorithm Neural Network, International Journal of Production Technology and Management (IJPTM), 13(1), 2022, pp 1–8

https://iaeme.com/Home/issue/IJPTM?Volume=13&Issue=1

1 INTRODUCTION

In recognition and control problems, the application of neural networks is increasingly popular, but due to the high nonlinearity of the activation function in neural networks, the implementation of neural networks in digital systems faces many difficulties and the accuracy

is not high high Besides, the design of the forward propagation process and the back propagation process of the network need to have a good coordination for the effective learning

of the neural network In this case, the neural network needs to be implemented on the hardware

platform [1-3] In fact, neural -networks can be implemented using both software and hardware Hardware-based implementations are especially useful in applications that require high speed and accuracy, such as in recognition technology and some other applications [4-5]

2 SURVEY OF KINEMATIC MODELS PERMANENT MAGNET

SYNCHRONOUS MOTOR

2.1 Overview of Permanent Magnet Synchronous Motor (PMSM)

The permanent magnet synchronous motor is composed of symmetrically distributed 3-phase windings and a permanent magnet-mounted rotor to create a gap magnetic field The elimination of the excitation circuit on the rotor side offers several advantages for PMSM such

as reduced copper loss, higher power density, reduced motor moment of inertia, and mechanically stable rotor construction.[2-5]

2.2 Kinetic model of PMSM

2.2.1 Equivalent Source

The power inputs for three-phase machines that are equal to the power inputs for two-phase machines have explanatory significance in measurement, analysis, and simulation

cs i cs v bs i bs v as i as v abc i

t

abc

v

i

(1)

i

p

: instantaneous input power

t

abc

v

: instantaneous phase voltage vector abc

cs

v

bs

v

as

v

,

, : phase input voltage a, b, c

abc

i : vector phase voltage abc

cs

i

bs

i

as

i

,

, : 3 phase instantaneous stator current

For balanced 3-phase machines, sequence currents are non-existent so the input power can be reduced:















ds i

r ds v

r qs i

r qs v

i

p

2

3

(2)

2.2.2 Electromagnetic Torque

Electromagnetic torque is the most important output variable for determining machine mechanical dynamics such as rotor position and speed It is derived from the machine matrix equation by looking at the input power and its other components such as resistive losses, mechanical power, and how much magnetic energy accumulates in the stator wire coil [6-7] The kinematic equations of PMSM can be written as:

      i

r G pi L i

R

Substituting [G] in the above equation we have the electromagnetic torque obtained as:

) ( 2

2

3

m N

r qs i

r ds i q L d L af

P

e

T =  + −  

2.2.3 Electromagnetic Torque

The dynamic equations of the PMSM in the rotor coordinate system can be expressed using flux linkages as variables Even if the voltage and current are discontinuous, the loop fluxes are continuous The stator and rotor loop fluxes in the rotor coordinate system are defined as:

r qs

i

q

L

r

qs =



af

r ds

i

d

L

r

The electromagnetic torque as a function of the loop fluxes is obtained by substituting the stator currents for the loop fluxes and is calculated as follows:







=







ds i

r qs

r qs i

r ds

P r qs

r ds af

q

L

P

e

2 2

3 )

1 ( 1

2

2

3

(7)

2.2.4 Electromagnetic Torque

The equivalent circuits of the PMSM can be derived from the stator equations and shown as Figure

Figure 1 Block diagram of a permanent magnet synchronous motor

3 CONTROL STRATEGY OF PERMANENT MAGNET SYNCHRONIC MOTOR

3.1 Structure of Drive System to Control Permanent Magnet Synchronous Motor

Figure 2 Schematic diagram of vector control structure of permanent magnet synchronous motor

torque control transmission system

Figure 3 The drive system regulates the speed of the permanent magnet synchronous motor

3.2 Control Strategies

The control strategies are considered in detail in synchronous motor drive systems:

1 Torque angle control is constant or shaft current d is 0

2 Power factor control

3 Controlling the total flux is constant

4 Angle control of total flux and complex current vector

5 Optimal control of torque

6 Loss reduction control on the basis of control in the maximum torque-speed range

Figure 4 Implement the control to keep the power loss constant

Figure 5 The drivetrain structure controls the minimum loss

4 INTRODUCTION TO ARTIFICIAL NEUTRAL NETWORK

Artificial Neural Networks are a powerful computing technology in the field of computer science With high-speed processing ability in learning solutions to a complex problem from a set of patterns, it has been widely studied and applied in recent years

Figure 6 Neural network model

5 RBF NEUTRAL NETWORK APPLICATION IN IDENTIFICATION AND CONTROL

Figure 7 Response of RBF-PI controller based on RBF neural network

Figure 8 Linear system recognition using RBF NN

u(k): control signal

Yp(k+1): linear system

Input of RBF NN includes: output of previous linear system + control signal

Figure 9 Tracking results with different learning rates (a) 0.1 (b) 0.25 (c) 0.5

From the results of the waveform, we can see that, with a small learning rate, the tracking response is slow The large learning rate leads to unstable response of the Neural network and strong fluctuations in the transition period

Figure 10 Simulation model for identification of PMSM drive system

Figure 11 The results of identification (a) and error (b) between the rotor speed and the speed

according to the Neural network

Figure 12 Output response result when following reference model with input command step for (a)

normal PI control and (b) RBF-PI control

Table 1 Comparison between conventional PI controller and RBF based PI controller

Figure 13 Simulation model for adaptive RBF-PI identification and control with PMSM drive

systems

Table 2 Input system initial values and PMSM parameters

Figure 14 Speed response by conventional PI controller and RBF-PI

We see that the rotor speed recovers at steady state when the system changes faster, the rotor speed is also closer to the reference speed

6 CONCLUSION

In the control algorithm, to adapt to the system uncertainty, RBF - PI is used to identify the nonlinear system and provide more accurate information for the parameters of RBF - PI The simulation results show that, according to the reference model, the speed of the PMSM can quickly and accurately respond after the proposed controller implements it simulation results achieved

Through the research process, we can see that the adjusted RBF - PI controller exhibits good control efficiency, improving performance compared to conventional PI controllers

That proves that the RBF - PI controller has good monitoring ability even though the system

is unstable and complex

REFERENCES

[1] A Esmaeili, N Mozayani, “Adjusting the Parameters of Radial Basis Function Networks Using

Particle Swarm Optimization,” IEEE International Conference on Computational Intelligence for Measurement Systems and Applications, pp 179–181, 2009

[2] S W Ellacott, J C Mason, I J Anderson, Mathematics of Neural Networks: Modes,

Algorithms, and Applications, New York, Springer, 1997

[3] H Robbins, D Siegmund, A convergence theorem for nonnegative almost super martingales

and some applications, New York, Springer, 1985

[4] F Sargeni, V Bonaiuto, “Digitally programmable nonlinear function generator for neural

networks,” Electronics Letters, vol 41, pp.143-145, 2005

[5] K Basterretxea, J M Tarela, I D Campo, G Bosque, “An experimental study on nonlinear

function computation for neural/fuzzy hardware design,” IEEE Transactions on Neural Networks, vol 18, pp.266–283, 2007

[6] S Himavathi, D Anitha, A Muthuramalingam, “Feed-forward neural network implementation

in FPGA using layer multiplexing for effective resource utilization,” IEEE Transactions on Neural Networks, vol 18, pp.880–888, 2007

[7] A Dinu, M N Cirstea, S E Cirstea, “Direct neural network hardware implementation

algorithm,” IEEE Transactions on Industrial Electronics, vol 57, pp.1845–1848, 2010