基于极限学习机算法的永磁机构动作时间补偿的研究

常规永磁操动机构的动作时间补偿的预测是采用加权平均值算法、多元线性回归法和人工神经网络算法(ANN)等对动触头的分/合闸时间进行估计和预测,但是加权平均系数的计算和线性回归系数的求解比较繁琐,而ANN网络具有训练速度慢、容易陷入局部极小点、学习率的选择难以确定等诸多缺点。研究了采用极限学习机(ELM)算法和BP神经网络算法,利用Matlab软件对永磁机构动作时间进行预测,通过对比分析,得到性能较好的算法。

Research of Action Time of Permanent Magnetic Actuator Based on Extreme Learning Machine Compensation

The action compensation prediction time of the conventional permanent magnetic operating mechanism is to use a weighted average algorithm, multiple linear regression and artificial neural network （ANN） algorithm, and many other methods of dynamic contact points and perform the time estimation and prediction, but the weighted average coefficient calculation and the solution of the linear regression coefficient are tedious, the training speed of ANN network is slow , and easy to fall into local minimum point, the learning rate is more difficult to determine, and many other shortcomings. This paper studied about the use of extreme learning machine （ELM） algorithm and BP neural network algorithm, using Matlab software to predict the movement time of the permanent magnet mechanism, by the means of the comparison and analysis, get better algorithm.