基于BP神经网络的输电线路雷击故障预测

雷击故障是输电线路故障的主要原因,需要对输电线路雷击故障进行预警以减少其造成的损失。传统雷区预报方法在预报准确性与雷区识别精度之间存在矛盾,难以进一步提高雷击故障预测的准确性。考虑到电力系统运行中积累了大量雷电和雷击故障的历史数据,本文建立了基于BP神经网络的输电线路雷击故障预测方法。首先在历史雷击故障统计分析的基础上,筛选出输入特征；然后分别应用粒子群算法和LM算法确定网络初始权值并进行网络训练；最后基于实际雷电数据和电网雷击故障数据对本文模型进行验证,仿真结果表明,本文方法能够预测80%的雷击故障,可为实际电网的雷击故障防御提供参考。#$NL关键词：雷击; 输电线路故障; 故障预警; BP神经网络#$NL中图分类号：请作者自查

BP neutral network based lightning fault prediction of transmission lines

Lightning fault is one of the main reasons of transmission line faults, which is necessary to be predicted to reduce the loss caused by lightning faults. The accurate detection of lightning area has conflict with the accurate lightning prediction, which makes it hard to further improve the accuracy of lightning fault prediction based on current method. With the development of date mining technology, large amount of measured lightning information, historical lightning faults information, which are deposited in the database of power system for many years, can be utilized to predict lightning faults. Therefore, a lightning prediction method for transmission lines is proposed this paper based on BP neutral network. Firstly, the input parameters of lightning faults are analyzed based on the historical lightning data. Then, a BP neutral network is employed to establish the prediction model, in which PSO is employed to calculate initial weight value and LM is employed to accelerate training speed of BP neutral network. Lastly, actual lightning information is adopted to verify the effectiveness of proposed method. Simulation results show that 80% of lightning faults can be predicted by proposed method which can provide a reference for power system operation.