基于遗传算法优化BP神经网络接触电阻预测.docVIP

基于遗传算法优化BP神经网络的接触电阻预测 孙海峰，沈颖，王亚楠 （华北电力大学 电气与电子工程学院, 河北 保定 071003） 摘要：接触电阻是反应导体间电接触性能的重要参数，在实际的工程中往往采用经验公式对接触电阻进行计算，精度难以满足要求。为解决这一问题，将遗传算法（GA）与BP神经网络相结合对接触电阻进行预测。通过实验得到数据，分别利用遗传算法优化BP神经网络、BP神经网络以及回归分析模型进行训练和测试，并将各算法所得误差进行对比。误差对比结果表明：遗传算法优化BP神经网络的收敛效果优于其他两种算法，且遗传算法优化BP神经网络所得接触电阻模型的相对误差平均值比BP神经网络减少了4.01%，比回归分析减少了4.72%，且预测效果较稳定。利用遗传算法与BP神经网络相结合的接触电阻预测模型较单独使用BP神经网络预测模型具有更好的非线性拟合能力和更高的预测精度。 关键词：电接触；接触电阻；遗传算法；BP神经网络；回归分析 中图分类号：TM769 文献标识码：A 文章编号：1001-1390（2019）00-0000-00 Prediction of contact resistance based on optimized BP neural network of genetic algorithm to optimize BP neural network Sun Haifeng, Shen Ying, Wang Yanan （(School of Electrical and ElectronicalElectronic Engineering, North China Electric Power University, Baoding 071003, Hebei, China）) Abstract: The contact resistance is an important parameter of the electrical contact performance between the reaction conductors. In practice, the empirical formula is often used to calculate the contact resistance, which is difficult to meet the accuracy requirements. In order To to solve this problem, the genetic algorithm (GA) combined with BP neural network is adopted to predict the contact resistance. The data are obtained Through through experiment,, the data are obtained , and the optimized BP neural network of genetic algorithm optimized BP neural network, the BP neural network and regression analysis model are respectively used for training and testing, and the errors obtained by each algorithm are compared. The results of error comparison show that genetic algorithm optimizes the convergence effect of BP neural network better than the other two algorithms, and the average relative error of the contact resistance model obtained by genetic algorithm optimization BP neural network is reduced by 4.01% compared with BP neural network, which is lower than the regression analysis 4.72%, and the forecasting effect is more stable. The contact resistance prediction model using genetic algorithm and BP neur