Research On Abnormal Voltage Quality Prediction Technology Of Distribution Network Based On Big Data And Machine Learning

With the continuous expansion of the distribution network,the operation of distribution transformer voltage and the situation of short-term and medium-term voltage crossing in the future have very important guiding significance for the optimal management of power distribution of power grid enterprises and the power use of power users.However,due to the existence of massive data at the level of big data in the vast distribution network,but due to the poor communication infrastructure of low-voltage distribution network,scattered support system,disordered data related to voltage analysis and governance,low adaptability of control technology,extensive management,failure to fully realize the hierarchical partition of distribution network,unified coordination of intelligent voltage optimization control system and other factors,the analysis based on big data is restricted This paper introduces the application of voltage out of limit early warning and optimal governance technology in distribution network.This paper first analyzes the core data requirements of short-term voltage overrun risk analysis and prediction of distribution network,and solves the problem of data collection and preprocessing for voltage risk analysis of distribution network.By studying the data fusion model and data fusion technology required by distribution network analysis,the data analysis technology based on distribution network data storage and preprocessing platform is proposed.Secondly,by studying the attribute analysis of distribution network based on topology structure and cluster analysis of distribution transformer equipment,a method is proposed to identify the connection between feeder and bus by using the correlation coefficient matrix of distribution transformer voltage and bus voltage under feeder.Thirdly,through the comparison of various schemes,the voltage risk prediction technology of distribution network based on the random forest classification prediction model is selected,which can reflect the influence of different factors on the voltage risk and provide the rectification direction for the optimal governance of distribution network voltage.Finally,after forming the theoretical basis of the system,the auxiliary system of voltage risk early warning and optimal governance is developed.The system can display the historical cross-border situation of distribution transformer in distribution network,and realize the risk warning of transformer cross-border in the next three months and the next six months.The calculation time of risk prediction is less than 60s.The auxiliary system has been successfully deployed in the big data platform of State Grid Yangzhou power supply company,which can better realize the risk prediction of distribution network voltage out of limit.