Research on Anomaly Detection Method for Electric Energy Meter Data Based on Autoregressive Continuous Hidden Markov Model

In order to improve the accuracy of anomaly detection in electric energy meter data, a method for anomaly detection in electric energy meter data based on autoregressive continuous hidden Markov model is studied. Adaptive filtering is employed to eliminate noise from electricity meter data. Kurtosis is selected as the indicator for abnormal features in the meter data, and these features are extracted. An autoregressive continuous hidden Markov model is then constructed to discern between abnormal and normal data, thereby achieving anomaly detection in electricity meter data. The experimental results demonstrate that this method can accurately and comprehensively extract the time-domain and frequency-domain features of electric energy meter data. The AUC value for anomaly detection reaches 0.96, and the false detection rate is below 5%, which can achieve accurate detection of abnormal data in electric energy meter data.