Analysis And Research On The Evaluation Index Model Of Electricity Stealing In Distribution Network

At present,the power system is gradually developing and stable,the smart grid is becoming more and more perfect,and the consumer electricity information collection system is widely used.With the application of the user electricity information collection system,a large amount of historical user electricity data has been accumulated.These data provide powerful data support for power sector grid planning,decision-making,consumer electricity behavior analysis,and load forecasting.In the power system,using big data analysis methods to find out the user's power consumption behavior to solve the problem of power theft is also a topic that has been ongoing.So far,the phenomenon of power theft still exists and is difficult to detect,which has caused the relevant departments of power companies to try their best and have little effect.The repeated prohibition of stealing electricity ignores the demands of ordinary people for normal life,destroys the harmonious atmosphere of the whole society,and increases the unstable factors at a certain level of the country.The existence of electricity theft not only brings about the huge loss of state property,but also damages the entire social order.At present,various new technologies have emerged endlessly,and the accompanying new technology stealing methods have also made it difficult to detect theft.In view of the above problems,this article summarize the abnormal power consumption caused by the theft of electricity.In order to fully reflect the characteristics of stealing electricity,comprehensively analyze the aspects of abnormal current,abnormal voltage,power factor,abnormal power,and line loss.Electric feature set.Then the method of feature selection is adopted,and the historical data is used to select the features of the sample data to complete the streamlined optimization of the features.Finally,based on the above characteristics,a BP neural network algorithm is used to establish a power stealing detection model.Experiments show that the final power stealing detection model can improve the accuracy of power stealing detection.The main research work of this article is:(1)A power stealing detection model based on the ReliefF feature selection algorithm and BP neural network algorithm was constructed,and then the real-life power stealing case data investigated and investigated in a certain area of Yunnan Province was used to analyze this detection model.Compare the accuracy of the feature-optimized theft detection accuracy.(2)Based on the analysis of the characteristics of power stealing caused by the power stealing methods in the distribution network,the basic principle of energy measurement in the power grid is derived,and the physical model of the energy meter measurement is given.Disturbance of the metering process makes the final metering of the electric energy under or uncounted.Secondly,it analyzes the current common methods of stealing electricity,including undercurrent,undervoltage,phase shifting,differential expansion,high-tech and other physical principles of stealing electricity.Summarize the abnormal power consumption caused by the theft of electricity.In order to fully reflect the characteristics of stealing electricity,comprehensively analyze the aspects of abnormal current,abnormal voltage,power factor,abnormal power,and line loss.Electrical feature set.(3)ReliefF feature selection algorithm is used to select the sample data containing the feature set of stealing electricity to obtain the features with high correlation with the results of stealing electricity.The features with low correlation are eliminated to form the feature index of stealing electricity.(4)Construct an evaluation model of electricity theft based on ABC-BP neural network.First,establish a BP neural network power stealing evaluation model,using optimized power stealing features and unoptimized power stealing features as model input,comparing the two sets of features when used as model input,network model performance and power stealing discrimination error rate,analysis It is concluded that when the characteristics optimized by ReliefF are used as the input value of the electricity stealing model,the electricity stealing model has good performance and the discriminating error rate of electricity stealing is significantly reduced.After that,an artificial bee colony(ABC)optimized BP neural network electricity stealing evaluation model is established.The optimized electricity stealing characteristics are used as model inputs to compare the final electricity stealing evaluation coefficient error rate of the BP neural network and the ABC-BP neural network electricity stealing detection model.