Research On Intelligent Operation And Maintenance Of Converter Station Equipment Based On Big Data Analysis

With the rapid development of smart grid and the increasing growth of power equipment,intelligent operation and maintenance has gradually become an important way for grid companies to improve production efficiency.The ultra high voltage direct current(UHVDC)converter station is an important part in the process of power transmission and transformation.The huge data generated during the operation of the converter station equipment has brought serious challenges to the traditional operation and maintenance methods.The intelligent management of the equipment must be implemented to realize the refined management.This paper takes a domestic converter station as the background,and conducts intelligent operation and maintenance research on the equipment operation and maintenance problems in the converter station.(1)In view of the main equipment and monitoring system of the converter station,the current status and requirements of intelligent operation and maintenance are analyzed from the perspective of the operation and maintenance data and operation and maintenance of the converter station.Then an intelligent operation and maintenance system framework based on the big data of the converter station equipment is proposed,including data source layer,data processing layer,feature analysis layer,model calculation layer and application of operation and maintenance layer.(2)Because of the single fault diagnosis method of equipment in the converter station,a method of equipment fault diagnosis based on data fusion is proposed.Multi-level equipment fault diagnosis models are established,including data-level fusion(data association fusion),feature-level fusion(primary diagnosis of multi-class logistic regression),and decision-level fusion(combined diagnosis of D-S evidence theory).Case analysis and method comparison show the superiority of the method.(3)To improve the imperfect early-warning algorithms for equipment faults in the converter station,a multi-granularity equipment early-warning algorithm based on LSTM and Seq2Seq is proposed,which combines the needs of operation and maintenance personnel with the sampling granularity,hour granularity,and day granularity to predict the state parameters of equipment.Then,combined with the fault warning criteria base,a matching analysis was performed for different types of faults.Finally,an example analysis of the converter device was performed and compared with the current alarm situation,which significantly improved the accuracy of the warning.(4)To remedy the problem of too many parameters in the equipment evaluation in the converter station,a method of state evaluation based on association rule mining and Bayesian Network is proposed.Mining the correlation of state parameters under various failure modes of equipment through association rules to improve the prior probability knowledge of Bayesian Network.After that,the self-learning of the Bayesian Network model is completed by using historical state information,current state information,and predicted state information.Then the distribution characteristics of the three types of information are considered to propose a method for evaluating the operating state of the equipment.Finally,the field operation data verification of the equipment in the converter station shows the effectiveness of the method.