Research And Application Of Power Intelligent Operation And Maintenance Based On Data Mining Technology

The original decentralized manual maintenance method of the power distribution system that relies completely on regular inspections by maintenance personnel fails to achieve a 100% fault discovery rate in the first time.Using the massive on-site operating data collected by the cloud platform to analyze the operating status of the power distribution system in real time helps to realize early warning of accidents,which will greatly reduce the negative impacts of power failures.Based on the data support of the existing cloud platform and data mining technology,this paper has carried out research on equipment operation and maintenance,energy management and electricity safety,which helps to realize predictive operation and maintenance.In terms of equipment operation and maintenance,in order to address the limitations of traditional methods to determine the temperature rise of transformer winding,this paper establishes a model for intelligent transformer maintenance based on correlation analysis.This model is able to estimate the normal range of transformer winding temperature rise under different load rate levels.It provides a universal intelligent maintenance method for transformers that have been put into use.In terms of energy management,this paper adopts k-means clustering algorithm to analyze the electricity consumption rules and conduct optimal scheduling of equipment electricity consumption.The clustering results help to make efficient and useful suggestions for off-peak power consumption management.Besides,this paper puts forward a method for short-term prediction of daily electricity consumption based on adaptive parameter similar day model.The parameters of the similar day model are trained from the sample data by using improved fruit fly optimization algorithm,which enhances the accuracy and versatility of the similar day model.In terms of electricity safety,in order to overcome the problem of insufficient model accuracy caused by inadequate consideration of influencing factors,this paper applies time series analysis to make ultra-short-term prediction of active power.The predicted value is compared with the real value,so as to determine whether an abnormal situation of power consumption happens.In addition,this paper establishes an adaptive leakage current protection model.The leakage current alarm value adjusts automatically according to the present leakage current level.This model reduces the number of false alarms caused by fixed leakage current alarm value.