Enhanced Ultra-Short-Term Load Forecasting Using MFO-Optimized LightGBM-Stacking

Ultra-short-term power load forecasting is crucial for ensuring the safe and stable operation of the power grid. However, due to significant fluctuations in power load, traditional forecasting models often struggle to deliver accurate predictions. To address this challenge, this paper proposes an ultra-short-term load forecasting algorithm based on MFO-optimized LightGBM-stacking. The algorithm first employs LightGBM to establish multiple base models for load forecasting, which are then combined into a base model layer for stacking model fusion. Another LightGBM model is subsequently introduced to integrate the base model layer, forming a comprehensive forecasting system. To further enhance prediction performance, the MFO algorithm is used to optimize the system's hyperparameters. Finally, case studies using actual load data from multiple regions demonstrate that the proposed method achieves higher prediction accuracy compared to traditional methods.