Short-Term Load Forecasting Based on Firefly Algorithm Optimized VMD-ATT-LSTM

This paper aims to address the issues of high load volatility, difficulty in determining model hyperparameters, and low prediction accuracy in new power systems by proposing an innovative short-term load forecasting model. The model integrates Firefly Algorithm (FA)-optimized Variational Mode Decomposition (VMD), Long Short-Term Memory (LSTM) network, and Attention (ATT) mechanisms, leveraging the strengths of each approach. First, the Firefly Algorithm is used to optimize key parameters of VMD, decomposing the raw load data into stationary subsequences to achieve denoising and effective decomposition. Then, the extracted features and decomposed subsequences are fed into the optimized Attention-LSTM prediction model for processing. Finally, the overall short-term load forecast is obtained by aggregating the predictions of each segment. Simulation experiments show that this method significantly improves the accuracy and stability of short-term load forecasting, demonstrating its advantages and application potential in low-carbon new power systems.