摘要
针对传统小电流接地系统故障选线方法准确度低、适用性差的缺点,采用改进遗传模拟退火算法优化BP神经网络(IGSAA⁃BP)的选线方法。该算法不但避免了传统BP神经网络对初始权值和阈值比重过高问题,还通过改进遗传算法的交叉变异概率公式和改进Metropolis准则来提高种群多样性,避免陷入局部最优。搭建配电网小电流接地系统仿真模型,获取零序电流的有功特征、5次谐波特征和暂态特征,输入IGSAA算法获得BP神经网络初始值,通过神经网络模型训练计算后获得选线结果。与其他BP神经网络算法对比表明,该方法在训练中收敛速度和复杂适应性较高,判断精度更好。
In allusion to the shortcomings of low accuracy and poor applicability of the traditional fault line selection method for small current grounding systems,a fault line selection method based on improved genetic simulated annealing algorithm optimized BP neural network(IGSAA⁃BP)is adopted.The application of the algorithm not only avoids the problem that the proportion of initial weight and threshold value of traditional BP neural network is too high,but also improves the population diversity and avoids falling into local optimization by improving the cross mutation probability formula of genetic algorithm and Metropolis criterion.The simulation model of the small current grounding system in the distribution network is built.The active power characteristics,5th harmonic characteristics and transient characteristics of zero sequence current are obtained.The IGSAA algorithm is inputted to obtain the initial value of the BP neural network.The results of fault line selection are obtained after training and calculation of the neural network model.The results of comparison with other BP neural network algorithms show that this fault line selection method has higher convergence speed,higher adaptability to complexity and better judgment accuracy in training.
作者
宋悦阳
黄洪全
陈延明
SONG Yueyang;HUANG Hongquan;CHEN Yanming(School of Electrical Engineering,Guangxi University,Nanning 530004,China)
出处
《现代电子技术》
2021年第22期15-20,共6页
Modern Electronics Technique
基金
国家自然科学基金资助项目(51567004)。
关键词
小电流接地故障
选线方法
单相接地故障
BP神经网络
模拟退火算法
仿真模型
small current grounding fault
line selection method
single⁃phase earth fault
BP neural network
simulated annealing algorithm
simulation model
