A Predictive Tracking Control Algorithm for Initial Angle of Brushless DC Motor Based on Reinforcement Learning

The initial angle detection and tracking control of brushless direct current motor (BLDC) is widely used in military, aviation, industrial, automotive, civil control systems, and household appliances. The motor is a nonlinear system, which is easily affected by external factors, so it is difficult to maintain the stability of the system with fixed control parameters. Therefore, this paper proposes a control method for initial angle tracking of BLDC based on reinforcement learning (RL). In order to determine the initial angle of the surface-mounted direct current (DC) motor, in the traditional Field-Oriented Control (FOC) algorithm, the pulsating high-frequency injection algorithm is introduced and combined with the RL theory. According to the given PI parameters, the appropriate rewards are obtained through the RL controller. Then the proportional integral (PI) parameters are dynamically adjusted to make the estimated angle closer to the true angle value. This method avoids the error of angle estimation caused by fixed PI parameters in different situations, and obtains the best results by dynamically adjusting the parameters. Finally, the simulation experiment is carried out to prove the feasibility of the proposed scheme.