A fast detection method for metering anomalies of three-phase energy meters based on sliding filter and decision tree

In the actual measurement of three-phase electricity meters, the presence of electromagnetic interference, environmental temperature, and other factors increases the noise in the measurement data, which affects the accuracy of anomaly detection. In this regard, a fast detection method for three-phase meter measurement anomalies based on sliding filters and decision trees is studied. Firstly, sliding filtering and sliding window matrix methods are used to reduce the dimensionality and filter the measurement data of three-phase electric meters. Then, a cascaded random forest is constructed using a CART decision tree, and the filtered three-phase meter measurement data is input into the cascaded random forest. The CART decision tree uses a binary method to partition the nodes. Finally, using the Gini coefficient as a measurement indicator, a cascaded structure is formed through layered stacking to output rapid detection results of outliers in three-phase electricity meter measurements. The experimental results show that this method can quickly detect measurement abnormalities of three-phase energy meters, timely detect meter flying anomalies and unexpected mutations, and improve the fault handling efficiency of three-phase energy meters.