Fault Analysis For Power System Based On Big Data

With the development of social economy,the demand for electricity is increasing day by day,and the distribution network responsible for delivering electricity to thousands of households is becoming larger and more complex.In the process,faults and anomalies in power grid become more and more frequent and diverse.Through the construction of smart grid,various sensors can continuously monitor the current,voltage and other indicators of power grid,and the distribution power grid has become a huge information network while transporting power resources.How to use automated data analysis to help the power grid to achieve more efficient maintenance has become an important topic and an important research direction.This paper analyzes the three most common types of faults and anomalies in the distribution network using various fault-related data generated by the distribution network,including current three-phase unbalance,excessive overload,and voltage anomalies.On the one hand,the analysis results can help the grid operation and maintenance personnel to predict the cause and severity of the fault,and arrange the maintenance plan more effectively.On the other hand,if there is a serious failure in the distribution network,it will bring huge economic losses.Afterwards,the repair is only to repair the sheep.Based on the automatic failure and abnormal analysis of big data,some abnormal lines can be found in time,and early warning of the occurrence of serious failure.The thesis mainly covers the following three parts:(1)Original data contains current,voltage,power,load and other time series data.It is high dimensional,so cannot be directly used as the input of clustering model.According to the needs of different types of faults and anomalies,this article extracts features and designs a series of characteristic parameters,and then use self-organizing competitive neural network to perform cluster analysis on these data of distribution network.These faults and abnormalities are clustered into more detailed classes.At last,a practical interpretation of the clustering results is given out.(2)To help power companies prioritize maintenance and repair work,it is necessary to prioritize the severity of faults and anomalies that occur in distribution network.In this paper,learning to ranking algorithm is used for distribution network fault analysis.Neural network and Rank Net algorithm are combined to realize the learning of the scoring function on the data set from labeled samples.Accurate ranking result which provided by this model can help maintenance department arranges their maintenance work more reasonably.(3)In order to improve the service of electricity companies,this paper analyzes the factors that affect the quality of service based on a large number of customer satisfaction data.This paper improves the decision tree algorithm based on Bayesian model.A satisfaction evaluation decision tree was established,and the decision tree was pruned to reduce overfitting.At the same time,cross-validation was used to judge the classification accuracy of the decision tree.The information gain rate of various factors was calculated to provide new ideas for electricity companies to improve service quality.