PI and PD Fuzzy Neural Network Controller Basedon Extended Kalman Filter for Brushless Drives

This paper presents development of PI and PD fuzzy neural network (FNN) controller for online speed tracking of brushless drives. This system is implemented by extended kalman filter (EKF) training algorithm to train PI FNN and PD FNN controller. FNN is a learning technique which finds fuzzy logic parameters by initiating techniques from artificial neural networks. Each FNN controller has four internal layers. Membership function and weights are modified according to the EKF training capability. The main objective is to replace classical PID controller by parallel PI FNN and PD FNN controller using EKF training algorithm. Parallel PI and PD FNN controller proves its improvement over conventional PID controller by comparing both learning algorithm. The hardware design is implemented with dSPACE DS1104 DSP and MATLAB. Results shows the superior learning capability and robust response of the proposed FNN controller in real time for different operating conditions.

I. Introduction

Brushless DC(BLDC) drives are widely used in industrial applications, electrical equipment’s and in so many fields because of its advantages of high efficiency, high power density, silent operation and so on. Brushless DC drives has a nonlinear and multivariable control system and this control strategy should be robust and adaptive. The classical PID controller has some disadvantages while controlling nonlinear and unbounded target. The two typical problems occur in PID response are 1) it overshoots initially as the reference step changes; and 2) it oscillates towards steady state.