Comparing Particle Filter, Adaptive Extended Kalman Filter and Disturbance Observer for Induction Motor Speed Estimation

Abstract

Electric motors in industry are required to operate at a certain speed with varying loads. In general, speed and position information can be measured using an encoder or tachogenerator on a motor shaft, but it will affect the cost and complexity factors. To reduce the cost factor and increase the reliability and robustness of the system, this information can be estimated, known as speed sensorless. This paper discusses three model-based estimation algorithms: Disturbance Observer (DO), Particle Filter (PF), and Adaptive Extended Kalman Filter (AEKF). The main topic in this paper is to evaluate these algorithms in estimating induction motor speed. Based on the performance testing results of the three algorithms, namely using root mean square error (RMSE) value, it was found that the DO algorithm is better than compared to the AEKF and PF algorithms.