Research On Power Consumption Behavior Detection Technology Based On Electric Power Oriented Iot Data

As the standard of daily life increases while being continuously affected by situations such as COVID-19,not only is there an increasing demand for electricity on the customer side,but also new patterns of electricity consumption behavior emerge,so it becomes important to detect customer's electricity consumption behavior to better suit their needs and to identify potential problems.At present,power IOT data is generated at the rate of millions every day or every hour,and there are many kinds of data,which contains important information reflecting customers' power consumption behavior,power quality,power equipment operation status,etc.Therefore,it is necessary to carry out power consumption behavior detection based on power IOT data.The user collection data originates from the smart meter on the customer side,which is the key data that best reflects the customer's electricity consumption behavior in the power IOT data.Therefore,this paper carries out the detection of users' electricity consumption behavior based on the data collected,and the analysis results can assist in the analysis of setting electricity tariffs and determining whether users have abnormal electricity consumption behavior(such as electricity theft and violation of electricity consumption).This paper introduces electricity consumption data into the detection of electricity consumption behavior and applies data mining techniques to all stages of the detection of electricity consumption behavior.The goal is to find a method that can make full use of the power IOT data and effectively classify and detect the electricity consumption behavior of customers.The specific research of this paper is as follows:1.To address the problem that the traditional power consumption behavior features do not consider the information contained in the power IOT data itself,data features are introduced into the power consumption behavior feature model,and a power consumption behavior feature set construction method is given that combines business-driven and data-driven features.The method uses business indicators to construct business features and extracts the frequency domain features of electricity consumption data based on the time-frequency transformation method.Using the two features together to characterize customers' electricity consumption behavior can provide a richer basis for electricity consumption behavior classification.2.A power usage behavior detection method for power IOT data is proposed.First,a feature importance assessment method based on maximum information coefficient and variance is given.Based on this method,business features and data features can be taken into account to obtain a more valuable set of electricity usage behavior features.Then,a recursive feature elimination algorithm-based electricity consumption behavior classification method is proposed,which uses the k-means-assisted DBSCAN algorithm with fixed parameters and combines the recursive feature elimination algorithm to find the best feature set,and achieves the classification results with advantages in terms of both cohesion and separation.After that,the user-side environment factor is introduced,and multiple algorithms are combined to predict the load and load interval of each type of users,and the electricity consumption behavior is detected based on the upper and lower limits of the load interval.Compared with the way of using a single prediction algorithm,the combination of multiple prediction algorithms is used,which integrates the advantages of multiple prediction algorithms and the prediction results have good stability.3.Designed and implemented the electricity behavior detection system based on power IOT data,the system consists of data management module,data processing module,electricity behavior detection algorithm library module,and result presentation module,providing functions such as data set selection,feature construction,adding algorithms,and result visualization,which can meet the needs of electricity behavior detection in terms of data access,data processing methods and result presentation.