Identification of abnormal electricity consumption of power users based on density clustering method

This study proposes a detection method based on density clustering for identifying abnormal electricity consumption of power users. Construct an outlier score calculation model for electricity fluctuation intervals through multidimensional feature analysis, standardize the data using Min Max, and quantify the degree of abnormal fluctuations based on density clustering algorithm. Input the calculation results into a weighted width learning model, combine with single hot encoding to construct a sample weight matrix, use ridge regression to optimize the output weights and achieve dynamic updates. This method effectively balances model accuracy and generalization ability through feature fusion and regularization processing, providing a new solution for anomaly detection in power data. The experimental results show that the proposed method exhibits a stable recognition accuracy of 95% and a low false alarm rate of $\mathbf{2. 1 \%} \mathbf{- \mathbf { 2. 7 \% }}$ in identifying abnormal electricity consumption, significantly improving the detection ability of abnormal behaviors such as electricity theft and equipment failure.