Research On Wire State Anomaly Prediction Method For Electric Power Centralized Monitoring Service

With the rapid development of electric power system, electric power equipment has accumulated a large amount of statistical data, part of the electrical equipment are in the collection and record the data associated with the electric net operating state at any moment, these statistics for power equipment in intelligent management accumulated valuable resources. At the same time, in recent years, the development of computer technology in the field of artificial intelligence are rapid,the application of machine learning technology is an important driving force for the development of it. Applying machine learning techniques to power data will provide a strong impetus to the development of intelligent power equipment.This paper is to study how to use machine learning technology to predict the abnormal condition of electric power equipment. Many previous work has made an important contribution to the anomaly detection, but the research work on the anomaly prediction, especially the abnormal prediction of the power equipment is not yet mature. On the one hand, this paper discusses the problem of how to predict the abnormal state of the wire by using the wire endpoint as the cut-in. The main research contents of this paper are as follows:Firstly, according to the relevant knowledge of power related to the wire endpoint data, we extract the corresponding feature attributes, and then the data are labeled to be abnormal or not. Different with "anomalies" in the field of data mining,the "anomaly" in this article is not judged by the data distribution, but according to the power specifications. Using power specification to carry out abnormal labeling is more relevant with electric power business, and can be more convenient and practical to use.Secondly, according to the characteristics of electric power data, using the machine learning algorithm, this paper sets up a set of abnormal warning model.The clustering and classification algorithms are used to model, in order to improve the model prediction accuracy and reduce the computational overhead, the sample sampling, feature selection, model fusion and so on are carried out.Finally, a set of abnormal warning system is designed according to the established model of abnormal predicting. According to the data collected by each minute of the electric power equipment, the abnormal warning model can predict the data after the completion of the data collection. In order to eliminate the impact of data drift, the system can be updated regularly or updated in accordance with the data drift dynamic.