Inertia Estimation for PMSM Drive System Using Artificial Neural Network

The model reference adaptive method (MRAM) has been widely used in the inertia estimation of the permanent magnet synchronous motor (PMSM) drive system. In this method, the deviation between the reference model and the estimation model is inevitable, which is the fundamental inducement for the estimation error. To this end, this work proposes an improved MRAM, which adopts a changing gain factor (GF) to reduce this deviation so as to enhance the inertia estimation accuracy. To provide the changing GF, this paper develops a single-neuron-based artificial neural network (ANN). It utilizes the deviation to adjust the GF dynamically. Furthermore, based on the instantaneous error-energy function, the proportional factor of the neuron is updated adaptively to force the estimated inertia to achieve a better tradeoff between stability and convergence rate. By simulations and real-time experiments implemented on the PMSM drive system under different working conditions, the effectiveness of the proposed methods is verified.

I. Introduction

Permanent magnet synchronous motor (PMSM) has gained extensive concern in industrial applications due to its excellent characteristics, such as large torque-to-inertia ratio and high efficiency. With the continuous improvement of the automation level, high-precision inertia estimation results are required for some servo control algorithms in the PMSM-based drive systems, including the self-tuning of the speed-loop controller [1],[2], the load torque observation [3], etc. It should be pointed out that imprecise inertia data will cause the undesirable control effects of these methods. Hence, it is essential to accurately estimate the inertia of the PMSM drive system.