PSO-Based HMLP Network Controller for Speed Sensorless Control of PMSM

Abstract A sensorless control of permanent magnet synchronous motor (PMSM) by using model reference adaptive system (MRAS) and artificial neural network is presents in this paper. The objective for this study is to improve the speed and position estimation of the PMSM rotor which used the MRAS with the conventional PI controller. In this study, the multilayer perceptron (MLP) network and hybrid multilayer perceptron (HMLP) network from family of artificial neural network (ANN) were evaluated. Before the controller can be used, the value of weight between the network layers need to be optimised. To train the controller weight, particle swarm optimization (PSO) is used. Finally, the proposed method is evaluated by comparing with the proposed controller in controlling the speed and position of PMSM. Simulation results under various speed and load conditions indicated that the PSO-HMLP network controller achieved well results than the compared controller in terms transient response and overall system analysis.