基于边界约束RBF网络的SRM磁链特性在线建模

Flux linkage characteristics on-line modeling of switched reluctance motor based on boundary constraints RBF neural network

针对开关磁阻电机(switched reluctance motor,SRM)本体非线性所带来的难以建模的问题,研究了SRM的磁链特性在线建模。在分析边界约束RBF网络(boundary constraints radial basis function,BC-RBF)的拓扑结构与学习算法的基础上,提出了以DSP为控制核心的SRM磁链特性在线建模方法,搭建了磁链特性在线建模实验平台。18.5k W和132k W两台样机的在线建模实验结果表明,该方法能够实现不同功率等级下SRM磁链特性在线建模,建模误差小于0.01Wb,该方法通过转速开环恒定电流方式在线检测获得电机原始数据,因而无需增加额外硬件检测设备,具有可行性和可移植性,为开关磁阻电机的磁链特性建模提供了一种易于实现的方案。

For the purpose of resolving modeling difficulties of switched reluctance motor （SRM）,brought by nonlinear characteristics,flux linkage characteristics on-line modeling of high power SRM was studied.On the basis of analyzing boundary constraints RBF neural network topology structure and learning algorithm,flux linkage characteristics on-line modeling method based on DSP was proposed.On-line modeling experiment platform of SRM was set up.Experiment results of 18.5 kW and 132 kW SRM show that the method can realize flux linkage characteristics on-line modeling under different power rating,the modeling error is less than 0.01Wb.The proposed method obtained original data by speed open-loop constant current on-line detection method,does not need additional hardware testing equipment and is feasible and portable for SRM modeling which provides an easy way to implement it.