Research Of Catenary Status Sensing And Anomalous Parameter Identification For Electrified Railway Based On Data-driven

With the acceleration of the transformation process of railways to electric in china,the construction of electrified railways is increasing.Catenary is a unique power supply network in the railway traction power supply system,which plays the role of transmitting electricity in the railway traction power system.The structural composition of catenary has similarities and differences with overhead transmission lines.The catenary has a specific voltage level,and its working environment is worse than that of the overhead transmission line.In case of fault,there is no backup line that can be activated,which will directly lead to the occurrence of an accident.Therefore,the research on the status sensing and fault parameter recognition of catenary is is of great engineering and social significance to ensure the safe and effective operation of railway transportation.This paper takes the catenary as the main research object,and focuses on the parameter failure of the catenary.Aiming at the problems that the current indicators in catenary status sensing system are incomplete and the fault parameters of catenary are difficult to identify,the work in this paper is as follows:(1)The research status of catenary status sensing and fault parameter recognition is summarized.First of all,based on the relevant domestic and foreign literatures in this research direction,starting from the development of electrified railways in recent years,the research significance of catenary status perception and fault parameters identification is expounded;Then,the research status of catenary fault parameters recognition is described in the dimension of method improvement and research status of catenary status sensing is expounded in the dimension of time advancement.Last but not least,the work has been done in this paper has been expounded and the chapter logical structure is shown by flow diagram.(2)The common methods of fault parameters identification and status sensing of electrified railway catenary are overviewed and two classical papers are reproduced.Firstly,the electrified railway catenary is introduced succinctly and a clear recognition of catenary is shown to the readers.Secondly,the common methods of catenary status sensing are introduced from the two steps of weighting and comprehensive evaluation.Then,the methods of catenary fault parameters identification based on neural network,image recognition and statistics are expounded respectively.Finally,two classic papers on catenary status sensing and fault parameters identification are reproduced,which is used to compare with the method proposed in chapters 4 and 5 of this paper,so as to reflect the superiority of the method proposed in this paper.(3)A method of catenary abnormal status detection based on least square support vector machine and FA-SVM is proposed.The catenary covers thousands of kilometers.So normal data and abnormal data of catenary are mixed together,which makes it difficult to distinguish normal or abnormal data.Aiming at the problems,an abnormal status detection method based on classification models in machine learning combined with swarm intelligence algorithms is proposed.Firstly,T-distribute stochastic neighbor embedding is used to reduce the original detection data to twodimensional data.Then,the least square support vector machine optimized by firefly algorithm is used to detect data after dimensionality reduction.Finally,the method proposed is compared with other methods.The results show that the method proposed can take into account the calculation time and detection accuracy,which has certain engineering practical significance.(4)A method of catenary status sensing method based on Grey-TOPSIS and normal cloud model is proposed.The current status sensing methods of catenary ignore some catenary indicators,select weighting coefficients artificially,calculate similarity from only one aspect and only consider the fuzziness of indicators without the randomness.Thus,a status sensing method of catenary with more complete indicators is proposed.Firstly,fuzzy analytic hierarchy process method and improved criteria importance though inter-criteria correlation are used to get subjective and objective weights,and least square method is used to get the united weights,which reduces the influence of artificial experience.Secondly,Grey-TOPSIS is used to obtain the weighted scores to each sensing object,which fully reflects the status of each sensing object.Thirdly,the normal cloud model is used to handle the weighted scores to obtain the membership degree of sensing objects in each status,which consider both the fuzziness and randomness.Finally,the principle of maximum membership degree is used to decide status of each sensing object,and the weighted average principle is used to verify the sensing results.The verification results show that the possibility of misjudgment of the status sensing method proposed in this chapter is small.The method provides a reference for the maintenance and repair of the catenary,and has practical engineering significance.(5)A method of catenary fault parameter recognition based on variational mode decomposition and NASFA-ELM is proposed.Catenary works as a key part in the electric railway traction power supply system,which is exposed outdoors for a long time and the failure rate is very high.Once failures occur,passenger travel and cargo transportation will be affected.Based on the above shortcomings,a method of catenary fault parameters identification based on nonlinear adaptive step firefly algorithm optimized extreme learning machine combined with variational mode decomposition is proposed in this paper.Variational mode decomposition is used to decompose the original data of catenary into a series of intrinsic mode function components,and the correlation coefficient method is used to filtrate intrinsic mode function components.The decomposed data are input into the nonlinear adaptive step firefly algorithm optimized extreme learning machine model to realize fault parameters recognition.Compared with extreme learning machine optimized by genetic algorithm and identification model without decomposition,the results show that the proposed model has better fault parameters recognition effect.Finally,the specific faults that may be caused by different fault parameters are expounded,which highlight the engineering significance of the method proposed in this chapter.