Research On Abnormal Power Consumption Detection Method Of Agricultural Smart Grid Based On Deep Learning

In recent years,with the transformation of the national power grid to the direction of intelligence,digitalization and informatization,the electricity consumption in the power market is getting larger and larger.For fast and accurate location of abnormal electricity consumption behavior in the power grid;it is important for timely detection of malicious electricity users,reducing economic losses and safeguarding agricultural production and development.How to use grid big data to effectively detect the abnormal electricity consumption behavior of agricultural users is a current research hotspot at home and abroad.However,the accuracy of existing machine learning algorithms based on artificial features is not high,and there is still room for further improvement.Deep learning,as an important extension of machine learning,is widely used in machine translation,sequence modeling and other fields.To revive the nation,the countryside must be revitalized.China’s comprehensive development of rural revitalization strategy,the issue of agriculture and rural farmers is a fundamental issue related to the country’s livelihood.Therefore,it is also better proof to develop rural agricultural power grid again.However,conventional abnormal electricity detection methods are highly dependent on professional knowledge and expert experience,making it difficult to achieve intelligence.In recent years,deep learning has been widely applied and can automatically extract and classify data features with high efficiency,which again provides valuable technical support for the research of abnormal electricity usage detection in agricultural smart grid,by using the most widely applied deep learning technology to detect abnormal electricity usage in agricultural smart grid based on the following two aspects.First: To improve the shortcomings of traditional convolutional neural networks in detecting abnormal electricity usage in agricultural smart grids,a 1D＿CNN-based model for detecting abnormal electricity usage in agricultural smart grids is designed to make it better for feature extraction of agricultural electricity usage data.A 1D＿CNN with a wide convolutional kernel is mainly designed so that the network can automatically learn the data features used for detection,and then batch normalization and Adam optimizer are introduced to optimize the network model.The input to the model is agricultural customer electricity consumption data,and a 1D＿CNN model is constructed to perform feature extraction on agricultural daily electricity consumption,and an experimental study is conducted to demonstrate the effectiveness of the proposed algorithm.Second: Since the 1D＿CNN detection model can only handle one-dimensional data of agricultural users’ electricity consumption,it is difficult to extract the periodic features of abnormal agricultural electricity consumption data and difficult to deal with the problem of time-series data,so,according to the characteristics of agricultural electricity consumption data set,we focus on designing a new convolutional layer,which can dig deeper into its periodic features,and use one week of agricultural electricity consumption data as the input of the network model.We construct a 2D＿CNN model for processing agricultural weekly electricity consumption data,add LSTM to 2D＿CNN to enhance the feature extraction capability for data with long-distance attributes,and build a 2D＿CNN＿LSTM-based model to detect abnormal electricity consumption in agricultural smart grid based on this.And experimental validation is conducted,which proves to a certain extent that the method has a good detection performance.