Research Of Low Current Grounding Fault Line Selection Method Based On Artificial Intelligence

The frequent occurrence of ground faults in distribution network has seriously affected the reliability of power supply.As the distribution network scales up,the conventional methods based on a single criterion is more difficult to meet the current route selection requirements.At the same time,when the system grounding fault occurs,the fault may further develop into a polyphase fault or arc grounding.The overvoltage generated by arc grounding will damage the device,which further worsens the fault and affects personal safety.Therefore,after a system fault arises,it is crucial from an engineering perspective to choose the fault line efficiently and conveniently.In this thesis,a simulation model of the small current grounding system is built using PSCAD/EMTDC.On this basis,transition resistance and arc grounding fault models are constructed,and various fault conditions are simulated by changing the relevant parameters.Based on the above model,the sum of the amplitudes of the five seventh harmonics are obtained by Fast Fourier Transform(FFT).The Variational Mode Decomposition algorithm(VMD)is taken to process the zero-sequence current of diverse scenarios,so as to achieve line energy.Finally,a power acquisition module is built to obtain the zero-sequence active power product component representing the fault trait of the line,so as to realize the extraction of the three types of fault characteristics of the line.At the same time,in order to accurately record the influence of different transition resistance,phase Angle and fault location on each electrical characteristic quantity of the model,a logic module was designed to record fault conditions,so as to accurately collect a large number of different types of fault characteristic data.Finally,based on the fault features,a BP neural network model is proposed to choose fault lines efficiently.Furthermore,considering the limitations of a single BP neural network,this thesis employs the whale optimization algorithm to optimize the BP neural network.On this basis,chaotic mapping equations and nonlinear inertia weights are utilized to enhance its global optimization capability and convergence speed.Finally,the characteristic data obtained by simulation were input into the BP neural network before and after optimization for line selection training,and the success rate of line selection before and after optimization and the error curve of the algorithm are compared.The simulation results demonstrate that the proposed method in this thesis can effectively improve the success rate of fault identification.