RBF neural network-augmented active disturbance rejection control for precise speed regulation of PMSM

In this work, permanent magnet synchronous motor (PMSM) speed control is studied using the Radial Basis Function (RBF) Active Disturbance Rejection Controller (ADRC). The proposed controller structure will adapt to momentary load changes and disturbances by using Radial Basis Function neural networks in conjunction with ADRC. This study demonstrates that RBF-ADRC can be superior to conventional controllers, such as model predictive control (MPC) and simple ADRC, for controlling PMSMs, particularly under dynamic load conditions. RBF-ADRC achieves a 93.7% decrease in settling time, a 92.9% reduction in rise time, zero overshoot in simulation under shaped-reference condition and a 80% decrease in steady state error as compared to sliding mode control (SMC). RBF-ADRC is superior to ADRC and SMC control of PMSM, not only from a mathematical point of view but also experimentally. The RBF-ADRC controller showed a 10.5% improvement in settling time, a 22.2% reduction in the rise time and a 50% improvement in steady state error with respect to ADRC control. The ADRC control system is effective at controlling motor speed across its operational settings due to the addition of RBF neural networks, which improve disturbance rejection and system stability.