Development Of Information Cleaning And Status Evaluation Platform For Main Transformer Equipment Based On Data Mining

As the hub equipment of power network energy conversion and transmission,the safety and reliability of oil-immersed transformer operation is directly related to the stable operation of the power system.With the improvement of power grid scale and voltage level,higher requirements are put forward for the reliability of substation equipment.With the continuous enrichment of condition monitoring equipment,the monitoring data gradually presents the characteristics of large volume,high dimension and multi-noise after the process of collection,transmission and storage.Therefore,based on the status index online monitoring data,this paper proposes the corresponding data processing and status evaluation strategy,and develops a system platform based on the system to verify the feasibility of the system scheme.In the aspect of data processing,this paper analyzes the characteristics of power equipment data,management technology and the existing problems and difficulties in the current data processing,and determines the data processing technology route with "screening out" and "restoring" as the main target.The on-line data are divided into two categories: core grounding current data and oil chromatography(DGA).For the current data,the idea of sliding window is introduced,and the dynamic outlier monitoring algorithm based on the time series autoregressive averaging(ARMA)model is proposed.Then,the Markov chain state transfer method is used to repair the data set to improve the data quality.For the DGA data,detailed will be piecewise linearization,the line segments are clustered based on the K-means algorithm to screen out the numerical points of different abnormal patterns in the data set,and propose a support vector regression(SVR)algorithm based on ameliorated particle swarm optimization(APSO),which improves the data set quality through the regression of the index state quantity.In the aspect of status evaluation,this paper based on the processed index online data set,the status evaluation system of online monitoring device and substation equipment is established respectively.In terms of the monitoring device,the relationship between the operation fault of the device and the representation of abnormal data is analyzed,the status level of the online monitoring device is set,and the running status of the device is evaluated by constructing fuzzy membership function.In terms of the status evaluation of substation equipment,a three-layer status evaluation system of target-data sources-index is established,and a processing method based on the idea of nearest neighbor is proposed for the serial index data to solve the problem that index data is difficult to process in the status evaluation.The case analysis shows that the designed state evaluation system can integrate the two indexes data of online monitoring and operation and maintenance,and give the evaluation results of power transformer and monitoring device respectively.On this basis,the substation equipment of data cleaning and state evaluation system platform is developed in this paper to realize the processing of condition monitoring data of the equipment,and provides the equipment status evaluation report.The research results have been successfully applied in An Electric Power Research Institute.