Research on current tracking error elimination algorithm for permanent magnet synchronous motor based on temporal convolutional networks

Current tracking error is a primary factor limiting motor control performance. This paper addresses this issue in model predictive control systems without integral controllers, using deadbeat predictive control as an example. Initially, the causes of current tracking error are analyzed to elucidate the fundamental reasons for its occurrence under model predictive control. This understanding is crucial for developing effective solutions. Leveraging the rapid convergence of Temporal Convolutional Networks (TCN), a TCN-based algorithm is proposed to eliminate current tracking error. TCNs are chosen for their quick adaptation capabilities, essential in dynamic motor control environments. By evaluating the motor's actual operating conditions, specific causes of current tracking error are identified. To address these swiftly, a voltage feedforward method is employed, allowing for rapid error elimination. This approach ensures the system can respond promptly to changes, maintaining precise control. Experimental results validate the proposed method's effectiveness, demonstrating its ability to eliminate current tracking error in model predictive control systems. Consequently, this significantly enhances the motor's load-carrying capacity at high speeds. The findings highlight the potential of the TCN-based algorithm to improve motor performance, marking a valuable advancement in motor control technology.