A Data-Driven Short-Term Line Loss Forecast Strategy Based on Multi-Head Autoformer Model

Line loss rate prediction is vital for efficient power distribution system management. With advancements in computer and artificial intelligence technologies, new methods for line loss prediction have emerged, however traditional data-driven methods like long short term memory (LSTM), autoregressive integrated moving average (ARIMA), and convolutional neural network (CNN) may not adequately represent real-world conditions due to idealized calculation conditions and data accuracy challenges. This paper introduces the Multi-head Autoformer algorithm, an enhancement of the classical Autoformer model for line loss rate prediction. Multi-head Autoformer algorithm exhibits superior performance in capturing intricate patterns and dependencies within line loss rate data than conventional strategies. Through real-world data analysis, this study demonstrates the effectiveness of the proposed method in accurately predicting line loss rates, addressing the challenges associated with power distribution system management.