【作者】 王建新；

【导师】 刘健；

【作者基本信息】 西安科技大学 ， 安全技术及工程， 2009， 博士

【摘要】 为了更好地开展接地网腐蚀故障诊断工作,确保人身、设备、电力系统和生产安全,本文开展了下列工作:建立了接地网支路电阻变化量与可及节点间电压变化量之间的非线性关系的增广故障诊断方程,并提出了一种采用迭代的方法逐步用线性方程来逼近非线性方程的迭代最小二乘接地网故障诊断算法。提出了一种基于禁忌搜索的接地网故障诊断算法,以可及节点间量测电压与估计电压偏差的平方和最小为适值函数,以支路电阻为解向量,分别以某个步长增加和减少支路电阻的策略进行邻域搜索。提出了一种基于蒙特卡罗模拟的接地网故障诊断算法比较和评价方法:在充分测试方案下,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,对迭代最小二乘算法、禁忌搜索法、单纯形禁忌搜索混合法、线性最小二乘法和非线性最小二乘法这几种接地网故障诊断算法的性能进行了比较和评价。提出了一种基于蒙特卡罗模拟的接地网故障诊断可测性分析方法:在合理范围内随机设置接地网各条支路的电阻生成样本,按照给定测试方案采用节点电压法计算得到测试数据并进行故障诊断,根据对各个样本诊断误差的统计分析,将相对误差和标准差小于给定阈值的支路被判定为明晰支路,其余支路被判定为不确定支路。在此基础上,提出一种采用最小生成树算法对明晰支路与不确定支路进行自动判别,并引入了补充明确顶点、支路二次分类、相近样本调整和理想样本自动生成等措施的改进方法,提高了判别正确率并减少了判别处理时间。提出了一种基于蒙特卡罗方法的干扰可测性分析方法,以反映量测误差对诊断结果的影响。将接地网故障诊断的可测性可分为结构可测性、测试可测性和干扰可测性3类,并定义了反映支路的不确定程度的指标。针对新建接地网,提出了一种基于遗传算法的监测极优化布置方案生成方法,能够以最少的监测极使接地网的所有支路都明晰,所提出的初始种群生成方法、交叉和变异方法能够避免不可行解,从而提高优化效率。建议了“抗干扰性最强原则”和“M+Y-X健壮性原则”对监测极布置方案的进一步优选。还探讨了在接地网撕裂诊断条件下的监测极布置方案优化问题。将上述成果应用到已建成接地网的测试中,提出了一种补充可及节点追加测试法,用以确定通过少量的开挖增加可及节点数量以使尽可能多的支路明晰的开挖方案。算例分析结果表明,所提出的方法是可行的和有效的。更多还原

【Abstract】 In order to diagnose the corrosion of the grounding grids and ensure the safty of power apparatus, electric power system and the human beings, the following contributions have been made:An augmented diagnosis equation to describe the nonlinear relationship between incremental branch resistances and incremental voltages between touchable nodes is established. An iteration least square approach is proposed, in which, an iterative algorithm is constructed to solve the nonlinear equation progressively by a series of linear least square approaches. A Tabu search based approach to diagnose the corrosion of grounding grids is investigated. The index is to minimize the summation of the squared errors between testing voltages and the evaluated values. The branch resistances form the solution vector. In each iteration step, a set of neighborhood is formed by increasing or decreasing the resistance of each branch with a certain level, respectively, while the other branches not changed.A Monte-Carlo based approach is proposed to evaluate and compare the performance of various grounding grid corrosion diagnosis algorithms. 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 of relative accumulated error of branch resistances and its standard deviations are used as the indexes for evaluation. Comparsions are made on several grounding grid corrosion diagnosis algorithms, such as, the iteration least square based algorithm, the Tabu based algorithm, the simplex and Tabu search hybrid algorithm, the linear least square based algorithm and the nonlinear least square based algorithm.A Monte-Carlo based approach is proposed to evaluate the testability of grounding grid corrosion diagnosis. The resistances of the branches of a grounding grid are set randomly to form many samples. According to a certain test scheme, the diagnosis results of branches of each sample are worked out. A stastical approach is carried out on the diagnosis results of all the samples, based on which, the branches with their relative errors and standard deviations less than the given threshold are determined as clear branches, the other branches are uncertain branches. A novel improved approach is suggested, which identifies the clear-branches and uncertain-branches automatically by a minimum spanning tree based algorithm and takes the measures of adding a clear vertex and an uncertain vertex, second classification of branches, similar sample adjustment and the ideal sample automatic generation. With the improved approaches, not only are the correctness consideablely increased, but also the amount of calculation needed in the testability evaluation is remarkably reduced. Besides, a Monte-Carlo based approach to evaluate the testability in case of the test data being with the measurent errors is also putforward. Three kinds of testability of grounding grids corrosion diagnosis are defined with their indexes, such as the structural testability, the measurement testability and the disturbance testability.As for the new grounding grids to be built, a genetic algorithm based method of layout optimization of monitoring lines is put forward. The optimum layouts of making all of the branches clear with the minimum number of monitoring lines can be worked out. The suggested ways of the initial population forming, crossover and mutation operation can avoid of the unfeasible solutions in the process, which ensures the efficiency. The best layout of monitoring lines is selected according to the proposed principle of the highest ability of anti-disturbace or the proposed robust principle of M+Y-X. The layout optimization of monitoring lines in case of splitting the grounding grid into small scaled sub-grids to test is also investigated. Applying above achievements into the diagnose of existing grounding grids, a new approach of adding touchable nodes and tests for grounding grid corrosion diagnosis is put forward, with the help of which, the positions of digging and adding touchable nodes with the least manpower to make certain uncertain branches clear can be obtained.The analyses of examples show that the proposed approaches are feasible and efficient.更多还原