Research On Detection Method Of Stealing Electricity Based On Load Data

With the vigorous development of China’s electric power industry,stealing electricity and other acts of stealing electricity are also plaguing electric power enterprises.How to ensure the normal use of electricity and maintain the economic benefit of electricity market is an urgent problem to be solved.Therefore,it is of great research significance to detect users’ behavior of stealing electricity and find them in a large number of users’ electricity consumption data.This article takes into account the characteristics of electricity theft data,based on the user’s daily load curve,yields a typical daily load curve.This article extracts the load rate,Daily peak-to-valley difference rate,Peak load rate,Trough load rate and flat period load rate in all 5 daily load curve characteristics,and uses entropy method to calculate the load curves of weight every day.In order to highlight the difference of electricity consumption between users,the discrete wavelet transform is used to decompose the typical daily load curve to get the trend characteristics of users’ electricity consumption,and the characteristics of users’ electricity consumption are combined with the characteristics of daily load curve.After obtaining the user’s power consumption characteristics,it is used as the basis of user clustering.By using the initial clustering center optimization k-means algorithm based on the spatial distribution density of samples,all power users are divided into four classes.Then,the distance between each type of user and the cluster center is calculated to judge the abnormal degree of the sample.Combined with the abnormal detection method of boxplot,three types of users stealing electricity,normal users and suspected users stealing electricity were matched and screened out.The final step is to detect the users who are suspected of stealing electricity.Using the Power System Analysis Synthesis Program(Power System Analysis Synthesis Program)referred to as PSASP,the modeling and analysis of electricity theft users,normal users and suspected electricity users,through the establishment of the user’s power consumption of the Station area model,to obtain the normal users,electricity theft users and suspected electricity theft users the maximum line loss rate,the line loss in the Taiwan area,and combined with the user’s electricity consumption peak,valley value,mean,peak,skew,skew seven characteristics to form the amount of electricity theft characteristics.The particle swarm optimization neural network(PSO-BP)algorithm is used to establish a thief recognition model,and the thief feature quantity is used as the algorithm input to further detect the suspected thief user.Experiments show that the method proposed in this paper has high accuracy and provides a new method for electricity theft detection.