The Online Parameter Identification Method of Permanent Magnet Synchronous Machine under Low-Speed Region Considering the Inverter Nonlinearity

Abstract

To realize the high-performance control of a servo system, parameter accuracy is very important for the design of the controller. Thus, the online parameter identification method has been widely researched. However, the nonlinearity of the inverter will lead to an increase in resistance identification error and the fluctuation of inductance identification results. Especially in the low-speed region, the influence of the inverter is more obvious. In this paper, an offline neural network is proposed considering the parasitic capacitance to identify the nonlinearity of the inverter. Based on the Kirchhoff equation in the static state of the motor, the nonlinear voltage equation is established, and the gradient direction of the weight coefficients has been re-derived. Using the gradient descent method, the identification error can converge to zero. Moreover, the d-axis voltage equation is established considering the nonlinearity of the inverter and an online adaptive observer was proposed. Based on the Lyapunov equation, the adaptive laws are derived. Further, the decoupling of the deadtime voltage and resistance voltage is realized by using the result of neural network identification. With the proposed algorithm, nonlinear identification of the inverter characteristics is realized, and the resistance and inductance identification accuracy in the low-speed region is improved. The effectiveness of the proposed methods is verified through experimental results.