Improved Performance for PMSM Sensorless Control Based on Robust-Type Controller, ESO-Type Observer, Multiple Neural...

Area of Science:

• Electrical Engineering
• Control Systems
• Artificial Intelligence

Background:

• Permanent magnet synchronous motors (PMSMs) are susceptible to performance degradation due to internal parameter variations (stator resistance, inertia) and load torque fluctuations.
• Classical control methods like PI controllers in Field Oriented Control (FOC) may struggle to maintain optimal performance under such disturbances.

Purpose of the Study:

• To develop and validate a robust control system for PMSMs capable of handling parameter and load torque variations.
• To enhance PMSM control performance by integrating a reinforcement learning (RL-TD3) agent with a robust controller.
• To evaluate the proposed control strategies against classical FOC using metrics including response time, speed ripple, and fractal dimension.

Main Methods:

• Synthesis of a robust controller designed to accommodate variations in PMSM parameters and load torque.
• Implementation of the robust controller and an enhanced version with a reinforcement learning twin-delayed deep deterministic policy gradient (RL-TD3) agent in MATLAB/Simulink.
• Development of four observer variants, combining neural networks and RL-TD3 agents, for estimating PMSM rotor speed and load torque.
• Comparative analysis of sensored and sensorless control strategies against classical Field Oriented Control (FOC).

Main Results:

• The robust controller effectively maintains PMSM control system performance across a range of parameter and load torque variations.
• The integration of the RL-TD3 agent significantly improved the performance of the sensored PMSM control system.
• Numerical simulations validated the superior performance of RL-TD3 trained agents in both sensored and sensorless control scenarios.
• Analysis indicated a higher fractal dimension (DF) of the rotor speed signal with the proposed efficient control systems.

Conclusions:

• The proposed robust control strategy, enhanced with RL-TD3 agents, offers superior performance for PMSM control compared to classical PI-based FOC.
• The developed observers accurately estimate rotor speed and load torque, contributing to enhanced control precision.
• The findings support the hypothesis that more efficient control systems lead to a higher fractal dimension of the controlled variable, indicating improved dynamic behavior.