Research On Fault Identification Of Through-phase In-phase Traction Power Supply System

Railway traction power supply system is very long,and the natural environment is complex.The fault happen easily on contact line,so the safety and stability of railway system are threatened badly,it is of great significance to identify the problem accurately and take action quickly.It is of great importance to find the areas where frequent failure,and the protection measures are important in advance.The safety of high-speed railway is easy to be disturbed by lightning strike,so it is necessary to study the new method of identify lightning failures,lightning interference and grounding fault.When lightning failures occur,the rapairer can make rapid response,prone to accurate grasp the areas of lightning failure,and make protective measures before failure occurs,or when failure occurs,the protection will not make false action.It will improve the stability of the railway.Co-phase traction power supply system is in the phase of research and development,it can solve the problem of unbalance and reactive power currents of electric railway by adopt single-phase ac-dc-ac converter.The main work is summarized as following:(1)Build up the AT power supply system base on direct feeding traction line,and build the lightning current model,analyzed the fault simulation.(2)The EEMD method combined with neural network for recognition of commutation failure is studied for identified lightning failuers,lightning faults and grounding faults.By calculating the energy,the correlation coefficient and approximate entropy of IMF component of the faults current signal linear mode component by the EEMD decomposition,then composing three feature vectors,using BP neural network and Elman neural network classifier to fault recognition and comparison.Compares the Fault recognition rate of three feature vectors and two neural network.Approximate entropy is the highest rate,and Elman neural network is higher than BP neural network.(3)Study the protection method of Extreme-point Sysmetric Mode Decomposition(ESMD)combined with support vector machine(SVM).By calculating the energy,the correlation coefficient and approximate entropy of IMF component of the faults current signal linear mode component by the ESMD decomposition,then composing three feature vectors,using support vector machine to fault recognition and comparison.The fault rate of support vector machine is higher than neural network.(4)According to the difference of high frequency and low frequency of fault,propose a fault identification method base on ESMD-SVM method.To detect faults near zero-cross point of voltage and normal fault.By calculating the approximate entropy of IMF component of the faults current signal linear mode component by the ESMD decomposition using support vector machine to fault recognition and comparison.The simulation result show that this method is effective.