Fault Diagnosis Of Grounding Grids Based On Intelligent Optimization Algorithm

Substation grounding grid is vital to ensure the operation safety of power systemas well as protect the operator and equipment. As grounding grid is buried in theground, the corrosion caused by soil acid may result in its grounding performancedeterioration, bringing accidents and huge economic losses. Generally, we use themethod of large-scale excavation to check the corrosion failure of grounding systems.This method is of much blindness, which is a waste of manpower and resources.Therefore, it is very necessary to analyze the fault diagnosis of grounding systemswithout excavation.In this dissertation, we analyze the grounding systems fault diagnosis, and carryon a thorough research on the model construction and methods of fault diagnosisbased on intelligence optimization algorithms and electrical network analysis.Based on previous research achievements, we create equivalent model ofgrounding grids through DC excitation. Based on this model, we apply RBF neuralnetwork to diagnose the grounding grids. Getting the test voltage of the diagnosis ofgrounding systems, the method takes the result of principal component analysismethod (PCA) as input samples. As it’s difficult to get the parameter of RBF neuralnetwork, we use the method of the difference algorithm (DE) to optimize the RBFneural network, to get an accurate diagnosis.The minimum energy principle was applied to establish a constrainedoptimization mode of fault diagnosis of grounding grids. Due to the difficulties inselection of penalty function, a new grounding network fault diagnosis methodcombines particle swarm optimization (PSO) with adaptive penalty function (APF)based on an adaptive penalty function is proposed. First, we do some statisticalanalysis of iterative population for the number of individuals that violate eachconstraint respectively. Then based on the proportion of feasible solutions, themethod is helpful for balancing the objective function and the penalty term effectively,avoiding the punishment being too heavy or too mild, which is helpful to make quickaccess to feasible solution area and search for the most satisfactory solution.Simulation results show that the algorithm has good diagnostic capabilities.For the diagnosis of large-scale grounding girds, according to the principle ofnetwork tearing, we shall tear the entire ground grid network into several sub-networks and free branches set, thus constructing a multi-objective optimizationmodel of sub-regional diagnosis. In order to overcome the blindness of selectinginitial value in Tabu Search Algorithm (TS), a solution combining differenceoptimized particle swarm algorithm (PSODE) with TS algorithm is proposed. Themethod utilizes PSODE’s preliminary optimization for subnets and free branches andthen using TS algorithm to determine the optimum solution. Simulation results showthat the proposed method generates more accurate result for fault diagnosis ofgrounding grids.