On-line identification and compensation-based model reference adaptive neural network speed control for LPMSM

Under the condition that the object model is definite, unvaried and linear, and that the operation condition and environment are ascertained, it is efficient to adopt a traditional control scheme. But for the situation of high precision and fine feed, the timing and variable factors such as the variation of structure and parameters of the object, the effects of all kinds of nonlinear factors, the changes of operation environment and environmental interferes must be considered to obtain a satisfying control effect. Nowadays modern control schemes are taken seriously in the study of LPMSM. In this paper, the authors propose a control scheme for the speed servo control of LPMSM with model reference adaptive and neural network techniques. Model reference adaptive neural network control is a new technique which is the combination of model reference adaptive control and neural network control. This kind of control method has double the advantages of the two methods. To improve the robustness of the system, the authors use an online identification technique to compensate the variation of the parameters and modify the calculation of the neural networks teacher value. The results of simulation show that the system has good speed servo performance, especially in overcoming the end effects of LPMSM.