LLA:Line Loss Anomalies Detection of Substation Areas Based on BACC Dual-Channel

Amidst the rapid advancement of the smart grid and expanding power system scale, line losses have emerged as a critical concern. This paper introduces a dual-channel model for detecting anomalies in station area line losses. The model adopts a dual-channel architecture constructed based on bidirectional Long Short-Term Memory (LSTM) and convolutional neural network (CNN), integrated with various-dimensional attention mechanism modules, aiming to capture global time series and local features separately. By concatenating these features and employing binary classification, the model accurately identifies anomalous line loss data in station areas. Experimental results demonstrate a detection accuracy of 98.43% across real-world datasets.