基于神经网络的开关磁阻电机转子位置估计

Rotor Position Estimation of Switched Reluctance Motor Based on Neural Network

位置检测是开关磁阻电机调速系统中的重要环节。实时、准确的位置信息是开关磁阻电机正确运行的关键。由于其转子位置角是各相磁链与电流的高度非线性函数,传统线性及解析的方法难以精确求得。该文基于神经网络并行处理及逼近任意非线性函数的特点,提出了基于神经网络的位置检测方案。借助于M ATLAB的神经网络工具箱,采用三种改进的学习算法对试验数据样本进行了离线训练,确定了用于位置检测的神经网络模型。为验证模型的有效性和准确性,对大量的数据样本进行了仿真。结果表明:该方法能够快速、准确地测量转子位置,鲁棒性和自适应性强。

Rotor posltion-detectlon is essential to the timing system of SRM. Duo to the doubly salient structure of SRM,its rotor position is a highly nonlinear function of stator windings current and flux linkage,so general linear methods are different to achieve precision results. In this paper.by utilizing the abilities of neural networks in parallel disposal and approaching discretional nonlinear function, the rotor position detection scheme based on neural networks is proposed. By adopting three kinds of improved neural network algorithms,the neural networks model is simulated for finding the rotor angle position at different currents from a suitable measured data for a given SRM. The data comprised flux linkage,current and rotor position. In order to testify the validity of the model,a lot of simulation was carried out. Results of experiments show that this scheme not only can acquire the rotor position timely and exactly,but has great robustness and adaptability.