Research On Winding Fault Diagnosis Method Of High-speed Railway Traction Transformer Based On ICA And Bayes Algorithm

With the continuous development of China's high-speed railway,its safe and reliable operation Is of great significance to promote the rapid development of social economy.The traction transformer is one of the important equipments to ensure the safe operation of the high-speed rail.The winding directly affects the transformation ratio of the traction transformer,which is decisive for the normal operation of the traction transformer.Therefore,the fault diagnosis of the traction transformer winding is also put forward higher requirements.However,the high-speed rail traction transformer is a special type of transformer with special voltage level,strong load change and frequent external short circuit.The winding of the traction transformer is always in a high load working state,so the accumulation of winding faults and the major faults are relatively frequent.Aiming at the above problems,this paper studies the fault diagnosis method of high-speed traction transformer winding based on Independent Component Analysis(ICA)and Bayes algorithm,and proposes Fast Independent Component Analysis(Fast ICA)algorithm for high-speed rail traction transformer.The winding fault diagnosis is carried out,and the Bayesian algorithm is used to optimize and improve the fault diagnosis method,which provides a theoretical basis for the winding fault diagnosis of the high-speed rail traction transformer to improve the safety and reliability of the traction grid and high-speed rail operation.Specifically include:(1)The topological structure of six high-speed rail traction transformers with single-phase,three-phase V/v,three-phase V/x,Scott,three-phase YNdll and impedance balanced wiring is studied.The simulation model of six traction transformer power systems is built up.The advantages and disadvantages of six high-speed rail traction transformers are compared.The common power supply modes of traction transformers are analyzed.The analytical model-based data-driven fault diagnosis method based on qualitative experience is analyzed.(2)The environmental noise separation method for fault diagnosis of high-speed railway traction transformer is studied,and the environmental noise separation model based on blind signal separation principle is established.Then the method of separating the environmental noise of the traction transformer voltage signal is analyzed.The calculation method of the ambient noise separation matrix of the traction transformer is studied.The preprocessing method of the voltage signal with environmental noise is studied.Then the Fast ICA ambient noise separation method based on the maximum negative entropy is studied.Finally,the feasibility and effectiveness of the Fast ICA ambient noise separation method based on the negative entropy maximum is proved by the simulation verification of the environmental noise separation method.(3)The method of Fast ICA algorithm for winding fault diagnosis of traction transformer is proposed.For the three-phase V/x wiring traction transformer winding fault,the I2 and squared prediction error(SPE)fault monitoring statistics are selected,and the fault contribution graph is proposed to locate the fault point of the traction transformer winding,which is completed by simulation.The initial diagnosis of traction transformer winding faults.(4)A fault diagnosis optimized method for traction transformer windings is proposed,which combines ICA and Bayes algorithm.The Ensemble I2(EI2)and Ensemble Squared prediction error(ESPE)comprehensive statistics are proposed to solve the winding faults.Monitoring improves the accuracy and efficiency of fault monitoring.Finally,the three-phase V/x wiring high-speed traction transformer is used as the platform.The feasibility and effectiveness of the improved algorithm combined with ICA and Bayes algorithm are verified by simulation.