Permanent Magnet Synchronous Motor Control Based On T-S Self Configuring Fuzzy Neural Network

This thesis takes the most widely used permanent magnet synchronous motor(PMSM)as the control research object.Aiming at the problem of poor dynamic performance of PID control used in most industrial control fields,the neural network can make the control system have better dynamic response ability through online learning and real-time correction of parameters.In this thesis,the speed controller of the motor is improved on the basis of the neural network.Aiming at the problem that the physical meaning of each neuron and parameter of the neural network is not clear and the existing engineering experience cannot be used,it is combined with fuzzy system which can store knowledge in rules to form a fuzzy neural network speed controller.The T-S fuzzy inference structure,which is simpler in mathematical analysis and is more conducive to combining with other control methods,replaces the commonly used Mamdani structure,and performs self-constructed feedback learning.Quick response and smooth operation under working conditions.Firstly,the existing surface mounted permanent magnet synchronous motor in the laboratory is taken as the research object,its mathematical model and motion equation are deduced,and the three-loop control simulation model of position loop,speed loop and current loop is established.Secondly,for the problem that the fixed parameters in the traditional PID control affect the dynamic response ability of the control system,the T-S fuzzy neural network is integrated into the PID control method to form a fuzzy neural network speed controller,and the optimal parameters are obtained through the online learning of the network to update the proportional,integral and differential coefficients of the PID speed controller,so as to improve the response ability and dynamic performance of the control system to changes in working conditions.For the deviation term of speed to controller output in fuzzy neural network parameter learning,the traditional learning algorithm can only use the symbolic function to approximate it.In this thesis,the RBF parameter identifier is used to collect the Jacobian information of PMSM,which improves the accuracy of the learning algorithm.The number of neurons has a great influence on the performance of the fuzzy neural network.In order to avoid the problem of over-fitting or under-fitting of the network caused by the improper selection of the number of neurons,it is proposed to modify the topology of the network through selfconstructured feedback learning,the structure learning and parameter learning are carried out at the same time,which speeds up the convergence speed of the network.Finally,in order to verify the effectiveness of the designed speed controller,the performance of PMSM under different working conditions is simulated and verified on Matlab/Simulink.Build an experimental platform with TMS320F28335 as the main control chip.Due to the limitation of hardware conditions,the fuzzy neural network algorithm cannot be fully reproduced.In the simulation,select a series of representative proportional,integral and differential coefficients to variable parameter PID control the motor to approximately restore the fuzzy neural network control.Through the comparison with the ordinary PID control,verifies that the T-S type fuzzy selfconstructed neural network speed controller has more excellent robustness and dynamic response ability.