Electricity Load Forecasting Based on DBO Optimized VMD Decomposition and Feature Screening

To overcome the power load data collection errors caused by the interference of external environmental factors, the existing algorithms fail to make full use of the data characteristics and lack accuracy problems. This paper proposes a combined VMD-MIC-Informer prediction model based on DBO optimization. The model optimizes the number of layers k and the penalty factor α of the VMD decomposition by the DBO algorithm to seek the best combination of parameters and subsequently calculates the correlation between each modal component of the decomposition and the factors influencing the electric load using the MIC method to enhance the capture of the characteristic variables. To evaluate the model, ablation experiments were conducted in this paper by measuring several prediction accuracy metrics (e.g., MAE, RMSE., and MAPE), where the MAE decreased to 24.40, the RMSE fell to 37.05, and the MAPE decreased to 0.28%. This indicates that the prediction model proposed in this paper has superior potential to effectively address the challenges of data noise and feature extraction, providing a powerful tool for improving prediction accuracy and stability.