Line Loss Prediction Based on Multi-dimensional Information Matrix and Multi-DAM-LSTNet

Line loss accounts for the most part of the energy loss in a low voltage distribution network. It is of great significance for the accurate prediction of the line loss ratio(LLR) in eliminating the abnormal faults of the transmission lines in time and ensuring the safe power supply. The existing line loss ratio forecasting methods rarely consider the seasonal trend data that affect the lines, and there is a lag effect when forecasting the non-stationary line loss series. Aiming at the above problems, a prediction based on the multi-dimensional information matrix and the multi-dimensional attention mechanism-Long- & Short-Term Time-Series Network (DAM-LSTNet) is proposed. Firstly, the maximum information coefficient (MIC) method was used to screen the distribution network characteristics and the seasonal trend parameters. Secondly, the variational mode decomposition (VMD) method optimized by genetic algorithm was used to decompose the historical line loss data to form a multidimensional information matrix with the characteristic parameters of the screening network. Finally, the multidimensional information matrix was put into the LSTNet network with the dimensional attention mechanism to predict the LLR. Numerical example analysis shows that compared with the existing methods, the proposed method has the advantages of sufficient consideration of the prediction parameters, adaptive change with the prediction weight, weak lag effect and high prediction accuracy. time series network; maximum information coefficient; variational mode decomposition; line loss. © 2024 Power System Technology Press. All rights reserved.