基于蒙特卡罗分析的接地网故障诊断算法评价

Evaluation of Grounding Grids Corrosion Diagnosis Algorithms Based on Monte-Carlo Approach

为了客观地认识和比较各种接地网故障诊断算法,基于蒙特卡罗方法,在一个60支路的接地网上,以完备的测试方案通过随机改变各支路电阻值和迭代初值的方式生成检验样本,以各支路电阻的相对累积误差均值和其标准差为评价指标,对迭代最小二乘法、非线性最小二乘法和禁忌搜索法三种接地网故障诊断算法进行了评价和对比。结果表明:迭代最小二乘法能够得到准确的明晰支路诊断结果,且与初值的选择无关;非线性最小二乘法和禁忌搜索法的诊断结果与初值的选择关系密切,若其初值的选择不恰当,则不能得到准确的诊断结果。

To evaluate and compare the grounding grid corrosion diagnosis algorithms, a Monte-Carlo based approach is carried out on a grounding grid with sixty branches under a complete test scheme. The testing samples are formed by randomly changing the resistances of branches and the initial values of iteration. The averaged value and standard deviation of relative accumulated error of branch resistances are used as the indexes for evalu- ation. Three algorithms are evaluated, such as, the iteration least squared approach, nonlinear least squares approach and Tabu search approach. It is concluded that the iteration least squared approach may obtain good diagnosis results for clear branches, which are not related with the initial values of iteration. Nonlinear least squares approach and Tabu search approach are much related with the initial values. Once a set of inadequate initial values are used, the diagnosis results may be with remarkable errors.