Current harmonics suppression strategy for motor drives using machine learning algorithm

• Main Author: Li, Tianyi
• Other Authors: Christopher H. T. Lee
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Permanent magnetic synchronous motor; Mechanical model; Current harmonics; Machine learning; Neural network; ADALINE
• Online Access: https://hdl.handle.net/10356/175973
• Institution: Nanyang Technological University

The Permanent Magnetic Synchronous Motor (PMSM) has been widely used in industrial field because of its simple structure, high accuracy, high efficiency and high power density. However, many disturbances existing in the motor control system may influence the overall performance, such as loading oscillations, parameters variation, friction force, temperature issue. Most of these factors may induce current harmonics, which will result in torque ripple and mechanical vibrations so that preventing the application in high performance occasions (i.e. robotics, Electrical vehicles). First of all, the existing solution to suppress harmonic current has been reviewed and divided to online compensation, offline compensation and periodical disturbance observer. The online compensation has high dynamic performance but usually need extra sensors to detect system parameters. Advanced periodical disturbance observe is effective and requires high computing ability of hardware. For offline compensation, it is hard for model-based method to adapt to different working conditions. Besides, considering the limitation of the existing strategies, the neuron network is known for its outstanding self-learning ability and excellent nonlinear approximation capacity. A neuron network-based method is proposed to suppress harmonic current. In motor control system, the dead time effect is considered as the main component to compensate. The outstanding dynamic performance of Adaptive Neuron Network (ADALINE) can perfectly cooperate with the steady performance of conventional PI controller. In the proposed method, PI controller and dual ADALINE is connected in parallel to suppress selected harmonics. Furthermore, the rationality of the proposed method is verified by the corresponding experiments. By analyzing the experimental result, the method is proved to be effective in steady, speed step, load transient, inductance transient condition, the performance of motor is optimized considerably. Also, based on the advantages drawn from the experimental result, the possible application for proposed method is briefly introduced which points out the prospect in industrial field.