Suppression of Periodic Speed Fluctuations for PMSM Using Improved Gradient Descent Algorithm

An AC disturbance suppression method for permanent magnet synchronous motor (PMSM) drives is proposed on the basis of a mathematical optimization algorithm. First, the AC disturbances of the system were equivalently moved to the controller input point (error feedback) according to the principle of shifting the introduction point of in a control flow diagram. Thus, additional harmonics with appropriate amplitudes and phase angles can be injected at this point to compensate for the effects of the AC disturbances. Second, a two-dimensional optimization method based on the gradient descent algorithm (GDA) was designed to obtain the injected harmonics with the optimal amplitude and phase angle. Furthermore, a reasonable evaluation function was designed to prevent the GDA from falling into a local optimum. Third, the iteration of the amplitude and phase angle was improved according to the evaluation function so that the two are approximately decoupled. This avoids the derivative calculations required by the traditional GDA and provides a better balance between the convergence speed and the accuracy. The method for selecting the initial value of the GDA iteration is also given, and the parameter robustness of the proposed GDA was analyzed. Finally, experimental validation was performed.