User Behavior Analysis Based On Power Marketing Data

As an important part of the power system,the power Marketing Department accurately identifies the abnormal behavior of users in the process of power consumption,which plays an important role in the safe operation of the power system.At present,the electric power Marketing Department uses the method of manual on-site inspection to detect the users with abnormal behaviors,which needs a lot of manpower and material resources.In view of the problems above,this paper proposes to use the historical consumption data of users to identify user behavior,establish a model at the data level,reduce labor costs and save resources.The original data used in this paper are user marketing data of Zhejiang Province in the past two years,and the data are analyzed from the aspects of time characteristics and spatial characteristics respectively:(1)Analyzing the time characteristics of user data,the long-short term memory add a back propagation network layer to increase compatibility of the time correlation characteristics of large span.The experimental results show that Bi-LSTM(Bi-directional Long Short-Term Memory)to user behavior identification accuracy than other kinds of networks,and low fluctuations,high stability.(2)Analyzing the space characteristics of user data,the monthly electricity to calculate the covariance between the data covariance matrix,then using CNN(convolution neural network)for feature extraction to get the users' feather sequence.The characteristics of sequence recognition model is established to fit characteristics and the results show that the model performed well in three kinds of user data.In addition,the abnormal behavior of the users in the platform area has a great influence on the line loss rate of the platform area,so this paper proposes a prediction model according to the line loss rate of the platform area to assist the detection of the abnormal behavior of users.This model uses the GAN for data expansion,and BP neural network for accurate prediction.This paper takes ensuring the safe operation of the grid system,and reducing the economic loss as a starting point to identify the abnormal electricity user behavior.Using Bi-LSTM and CNN respectively from aspects of time and space characteristics to analyze the model and forecasting area line loss to assist the user behavior detection.The experimental results are valuable for user management of marketing system.