架空输电线路档距段雷击跳闸预测

Lightning trip predictionin span section of overhead transmission line

为解决当前输电线路雷击跳闸预测模型考虑因素不全面、预测不准确等问题,本文通过分析输电线路雷击档距段雷击跳闸预测存在的问题,介绍模拟退火遗传算法优化的BP神经网络算法的优点,并进行档距段雷击跳闸预测。本方法具有模拟退火算法避免陷入局部极小值的特点,并使用遗传算法并行化高效搜索最优解。解决了BP神经网络在模型预测中存在的训练效率低下、收敛值易为局部值的缺点。本文实验首先进行数据的归一化处理、数据集的选取、设置算法相关参数,将贵州省某电网的杆塔的台账数据和雷电监测数据作为源数据进行训练分析,以误差标准和性能标准衡量算法优劣,并经测试样本验证。通过实验结果的对比表明本文提出的算法在档距段雷击跳闸预测上更准确。

In order to solve the problems of incomplete consideration and inaccurate prediction in the current lightning trip prediction model of transmission line,this paper analyzes the problems existing in the lightning trip prediction of transmission line span section,introduces the advantages of BP neural network algorithm optimized by simulated annealing genetic algorithm,and carries out the lightning trip prediction of span section.This method has the advantage of simulated annealing algorithm avoiding falling into local minimum,and genetic algorithm is used to search the optimal solution in parallel.The disadvantages of low training efficiency and local convergence value of BP neural network in model prediction are solved.In this experiment,firstly,the data are normalized,the data set is selected,and the relevant parameters of the algorithm are set.The tower account data and lightning monitoring data of a power grid in Guizhou Province are used as source data for training and analysis.The algorithm is evaluated by error standard and performance standard,and verified by test samples.The comparison of experimental results shows that the proposed algorithm is more accurate in predicting lightning trip-out at span.