摘要
为解决接地网腐蚀支路诊断中牛顿-拉夫逊算法呈病态、受初值影响较大与DE算法求解随机、易陷入局部最优解的问题,提出了用于接地网腐蚀支路诊断的DE-MNR混合求解算法。首先借助于电网络理论提出了基于MNR算法与DE算法的接地网故障诊断模型;在此基础上给出了DE-MNR诊断模型及具体的求解流程,该模型首先采用DE算法在支路电阻分布空间内搜索出较优的电阻分布,然后将其输出给MNR算法,MNR算法在具有较优初值的条件下快速收敛至最优解。60支路和112支路的接地网仿真实验表明DE-MNR算法能准确有效的诊断出接地网的多支路故障并判断各支路的腐蚀程度;通过与MNR算法的对比进一步说明了所提算法的有效性及稳定性。
To solve the problem that the Newton-Raphson algorithm is ill-conditioned and greatly af⁃fected by the initial value,and the DE algorithm is random and easy to fall into the local optimal solution in the diagnosis of the corroded branch of the grounding grid.This paper proposed an improved method,DE-MNR algorithm,for the diagnosis of the corroded branch of the grounding grid.First,with the help of electrical network theory,the grounding grid fault diagnosis models based on MNR algorithm and DE algorithm are given.Then the DE-MNR diagnosis model and specific solution process are given.The model first uses the DE algorithm to search for a better resistance value distribution in the resistance value distribution space.Then the MNR algorithm quickly converges to the optimal solution with the better re⁃sistance value distribution calculated by DE as the initial value.The 60-branch and 112-branch grounding system simulations show that DE-MNR algorithm can effectively and accurately diagnose the multi-branch fault of the grounding network and judge the corrosion degree of each branch;the compari⁃son with the MNR algorithm further illustrates the effectiveness and stability of the proposed algorithm.
作者
李晨旭
范丽君
黄浩
李准
王昕
LI Chenxu;FAN Lijun;HUANG Hao;LI Zhun;WANG Xin(School of Electrical Engineering,Shanghai Electric Power University,Shanghai 200090,China;State Grid Jiading Power Supply Company of Shanghai Municipal Electric Power Company,Shanghai 201800,China;Center of Electrical&Electronic Technology,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《电瓷避雷器》
CAS
2024年第1期103-110,共8页
Insulators and Surge Arresters
基金
国家自然科学基金项目(编号:61673268)。
关键词
接地网
腐蚀支路诊断
电网络理论
DE-MNR算法
grounding grid
corrosion branch diagnosis
electrical network analysis
DE-MNR algorithm