Study On The Condition Assessment Of Power Grid Equipment Based On Big Data Analysis

Power transformer plays a very important role in power system,and its safety and reliability are fundamental to the stability of the power grid system.Using online monitoring techniques to do fault prediction and diagnosis is a meaningful topic in intelligent power system study,which can greatly reduce the power grid's operation risk and maintenance costs by taking appropriate actions to prevent the heavy accident.However,as the development of power grid big data,the raw data from a single online monitoring system is now rarely sufficient for training and the poor data quality can rarely meet the precise analysis requirement for data noise,data absence and data redundancy.This paper proposes a data preprocessing scheme oriented to transformer fault diagnosis,which include two parts: a missing data prediction model based on improved support vector machine(OSVM)and refined support vector machine(RSVM),and a fault classification framework based on correlation analysis and principal component analysis.Through experimental validating,our method shows better performance than the traditional neural network and SVM to predict the missing data,improving the data quality of the raw data.Besides,when applied to fault diagnosis,our method also shows a higher accuracy and less computing time and introduces oil temperature as a new fault classification index.What's more,the proposed scheme works better and better with the increase of data dimensions,indicating that our method is worthy of further promotion and research to meet the big data times.