基于改进BP神经网络非正弦供电下异步电动机转速估计

非正弦供电下异步电动机电压、电流和功率都比较难以确定，电压、电流波形发生严重的畸变，含有大量的谐波，对于非电气量的转速易受到大量谐波的干扰，脉动比较严重，准确地测定比较困难，传统的测速方式成本高，需要维护，不经济，但对于电动机的电压、电流容易通过PT和CT获取，易实现转速的在线估算和测量，本文提出了改进的BP神经网络来训练非正弦供电下异步电动机机端电压、电流、功率因数和转速之间的复杂映射关系，并了改进的BP算法，仿真结果和试验结果证明了利用此神经网络的软测量技术来估计转速达到了较高的精度。

Estimation of Rotation Speed for Induction Motor under Non-sinusoidal Power Supply Based on Modified BP Network

It is difficult to measure the voltage, current and power for induction motor under non-sinusoidal power supply, and this results in distortion of the wave of voltage and current which includes many harmonics. It is not easy to accurately measure it because these harmonics interfere with the non-electrical rotation speed and cause serious fluctuations. The common method involves high cost and needs maintenance, but the voltage and current of induction motor are so easy to gain through PT and CT that the rotation speed can be estimated and measured on-line. The paper presents the modified backpropagation neural network to train the complex mapping connection between the voltage, current, power factor and rotation speed for induction motor under non-sinusoidal power supply, and introduces the modified backpropagation algorithm. Results of simulation and experiments show that it is feasible to use the soft measurement technology of the neural network to gain higher accuracy.