Decision Tree based Electricity Theft Detection in Smart Grid

One aspect of using smart meters is detecting anomalies in advanced metering infrastructure. Electricity theft, as a well-known anomaly, can be discovered by various machine learning algorithms. In this paper, decision tree, random forest, and gradient boosting methods are implemented and performed on collected power consumption data from 114 single-family apartments to detect non-technical loss, which are generated by various scenarios. Performances of these algorithms are analyzed with and without clustering on the measured data of each user.