Active Disturbance Rejection Control of Bearingless Permanent Magnet Slice Motor Based on RPROP Neural Network Optimized by Improved Differential Evolution Algorithm
Chaobin Sun, Huangqiu Zhu
Abstract
In order to realize the strong antidisturbance capability and the precise control of suspension forces in a bearingless permanent magnet slice motor (BPMSM), an active disturbance rejection control (ADRC) strategy based on the combination of resilient backpropagation neural network (RPROPNN) and improved differential evolution (IDE) algorithm is proposed. Based on the mathematical model of the BPMSM, the first-order ADRC controller of the rotating part and the second-order ADRC controller of the suspension part are designed, respectively, for the BPMSM according to the different orders of the BPMSM system. Then, the elite group bootstrap mechanism and parameter adaptive techniques are introduced to improve the DE algorithm to speed up its convergence and enhance the global search capability, and the initial weights of RPROPNN are optimized using the IDE algorithm to obtain the best network model. According to the feedback information of the BPMSM, the optimization mechanism and self-learning capability of RPROPNN are used to adjust the parameters of ADRC to reduce the dependence of ADRC on parameters. The comparative experimental results indicate that the proposed ADRC optimization design method has stronger robustness when disturbances occur, and the method is feasible and effective.