State Information Mining And Fault Diagnosis Of High Voltage Circuit Breakers In Big Data Environment

With the rapid development of smart grids,the power industry has entered the era of big data with the information society.In the field of on-line monitoring of power equipment,with the development of sensor technology,it is faced with new problems such as diversified types of monitoring data and dramatic increase in data volume.After the data is abundant,data with low “cost-effectiveness” is extracted.This has led to a series of problems like the difficulty in extracting effective features,the difficulty in accurately determining the state of equipment operation,and the difficulty in providing a panoramic display of the power equipment operation.In order to deal with the difficulty of extracting effective features from mass data,deep learning technology is introduced into the feature learning domain for high voltage circuit breaker monitoring.Combined with the characteristics of data samples,the Deep Belief Network(DBN)is used to deepen the feature mining of data.Based on the theoretical analysis of DBN,a greedy training method is proposed to achieve the purpose of deep learning of data characteristics and guarantee the efficiency of network training.For circuit breaker fault diagnosis,including fault trend identification,this paper combines DBN with classifiers,and uses the deep abstract features extracted from DBN as training samples for classifiers,and then trains deep classification networks to solve pattern recognition problems.By collecting data samples through fault simulation tests,the deep classification network established in the article is trained and tested,and excellent results are obtained.At the same time,it also compares with the Back Propagation(BP)neural network and Support Vector Machine(SVM)classification results,which further proves the superior performance of deep learning technology.Aiming at the problem of panoramic display of circuit breaker operation status,based on the successful application of deep learning technology,this paper designed a high-voltage circuit breaker condition monitoring and panoramic display platform under the background of power big data,and proposed to expand to any high voltage device.First,a database for power big data is established as a data sample for the diagnosis of equipment operation status.Then,data mining and feature learning modules,decision guidance and visual display modules are merged,and the overall architecture of the panoramic platform is proposed.deep learning technology is introduced to avoid the problems that frequently used intelligent algorithms are easy to fall into the local minimum,do not adapt to large amount of data,and need to manually extract feature parameters.The excellent performance in the experiment also proves the significance of the method.This paper preliminarily designs a panoramic display platform under the background of power equipment big data,providing some ideas for future research work.