Memoryless radial basis function neural network based proportional integral controller for PMSM drives

Abstract

This research paper presents the memoryless radial basis function neural network (RBFNN) dependent proportional integral (PI) controller for permanent magnet synchronous motor (PMSM) drives. The proposed RBFNN is memoryless adaptation scheme since the algorithm neither use the past samples nor retain the previous adaptation direction and hence the observations are used only once. Firstly, mathematical model for PMSM drive is explained and then structure of RBFNN controller is formulated. RBFNN consists of single input, a hidden layer with five neurons, a single output layer that are used to track the speed of the machine. The Jacobian matrix obtained from RBFNN is used to adjust the gain of the speed and applied to PI controller. The performance of the PI based vector controller with and without RBFNN are compared through the simulation by using MATLAB/Simulink environment. The simulations show that the proposed RBFNN outperform the existing PI based control method and enhances the speed tracking of PMSM drives.