Grounding Line Selection Resonance Distribution Network Failures

The quality of power supply is closely related to the safe and reliable operation of distribution network. In China, The6-66kV distribution system is often connected to ground through arc-suppression coil. When single-phase ground fault occurs, the inductive current of the coil would compensate the capacitance to ground current of the system, reducing the fault current which flows through the grounded point. Due to the character mentioned before, when ground-fault happens in resonant system, it could continue to run for1-2hours according to the regulations. Thus, the coil could improve the reliability of power supply, but this would increase the difficulty in fault line selection at the same time. With the development of distribution network, the application portion of the cable line and cable mixed lines increase with enlarging the capacitance to ground current. The long time running in such condition, the single-phase fault of the system will develop to two points or multiple points fault and the arc grounding may damage the safe operation and usage life of the electrical apparatus. Therefore, a novel fault line selection method which could adapt the practical scenario is needed urgently.Large amount of fault line determination approaches have been proposed by many scholars, where the dominant method is still the’comparison of amplitudes and phases in group’. With the development of the processing methods of digital signals, the time-frequency characteristics differences of the zero-sequence current between sound feeders and fault feeder are extracted to form the fault line selection criterion. As the artificial neutral network (ANN) evolves, more and more intelligent methods come into being. A resonant system model is built by MTLAB/Simulink in this paper. Based on this, the transient process of the single-phase fault is analyzed and two algorithms are proposed by analyzing the5ms zero-sequence current data of each feeder after fault. Firstly, as for the common unreal-ground phenomenon in resonant system, a method based on the energy ratio of the high and low frequency wavelet coefficients extracted from zero-sequence voltage to distinguish this from fault is proposed. Secondly, two new faulty line selection methods are approved.one is an intelligent method which utilizes the RMS values of the continuous wavelet transformation coefficients in characteristic band extracted zero-sequence currents combines ANN. No explicit criterion need to be presented in this method, it just need input the RMS values obtained in characteristic band into fault line selection ANN, and output the fault line serial number. Train the ANN model by using large amount of samples to detect the fault line. Another method is to employ one order straight line fitting with the wavelet composition coefficients of zero-sequence currents in characteristic band. Form the criterion by the signs of the lines extracted by sign function. Large amount of simulations show that the method is accurate and reliable.