Abnormal Detection Method Of Smart Grid Based On Deep Learning

With the advent of the era of informatization and intelligence and the continuous upgrading of power systems,the security issues of smart grids are facing severe challenges.Among them,the legal and safe use of electricity by power users determines whether the power grid can operate safely and stably.Therefore,it is of great significance to study abnormal electricity use detection methods for smart grids.In the process of abnormal electricity consumption detection,effective extraction of data features is the key to detection.However,the traditional abnormal electricity consumption detection method requires a large amount of professional knowledge and expert experience when performing feature extraction,and it is difficult to realize the intelligentization of abnormal electricity consumption detection.With the development of deep learning,deep neural networks can automatically extract and classify data feature information quickly and effectively,which provides new ideas for the study of abnormal electricity use detection in smart grids.This paper uses deep learning methods to detect abnormal power consumption in smart grids.The specific research content is as follows:(1)Aiming at the low accuracy of the traditional convolutional neural network for abnormal power consumption detection,a smart grid abnormal power consumption detection model based on 1D＿CNN was established to make it better to extract features from daily power data sets.Mainly designed a one-dimensional convolutional neural network with a wide convolution kernel,so that the network can automatically learn the data features for detection,and use batch standardization and Adam optimizer to optimize the network model.Taking the user’s daily electricity load data as the input of the network model,a onedimensional convolutional neural network model for processing daily electricity data is constructed.Experimental results show that this method improves the detection performance of abnormal electricity consumption.(2)Aiming at the problems that the 1D＿CNN detection model can only process one-dimensional daily electricity consumption data,it cannot extract the periodic characteristics of abnormal electricity consumption data well and it is difficult to process time series data.A smart grid abnormal electricity consumption detection model based on 2D＿CNN＿LSTM is established.Combining the characteristics of the smart grid electricity data set,the convolutional neural network model is improved again,and the one-dimensional daily electricity data is converted into two-dimensional weekly electricity data on a weekly basis.A special convolutional layer is designed to analyze the data.Periodic characteristics are extracted at a deeper level.With the user’s weekly electricity load data as the input of the network model,a two-dimensional convolutional neural network model for processing weekly electricity data is constructed,and it is added on the basis of the two-dimensional convolutional neural network.Long and short-term memory network to improve the feature extraction ability of long-distance attribute data.Experimental results show that this method has better detection performance.