A SVM-Based Fraud Detection System Using Short-lived Electricity Consumption Data

In recent years, advanced smart meters have been widely used in Smart grids, which monitor electrical power consumption over fine-grained time intervals and has made it easier for electricity companies to monitor anomalies in the network. However, smart meters are subject to many security vulnerabilities. A short-term fraud detection method based on the Support vector machine (SF-SVM) is proposed in this paper. The method only needs to collect and store a small amount of the user's recently electricity consumption data to detect problematic users. Using a small amount of data can reduce data storage and reduce the cost of data remote transmission. Furthermore, user privacy can be better protected. The system automatically collects the electricity consumption data of the grid and users at a certain period. When the system detects that the difference between the amount of electricity provided by the regional grid and the amount of electricity consumed by the users exceeds a threshold, the system changes to a suspicious state, and triggers fraud detection process. The system introduces machine learning algorithms to extract features from users data, and finally find suspicious users. Simulation results show that the method effectively detect abnormal users.