Electricity Demand Forecasting Using Transformer-Based Model Optimization and Interval Calibration

This study proposes a power load prediction framework based on multiple machine learning models, including Naive 7, XGBoost, Random Forest, and a deep learning model based on Transformer. The framework integrates hyperparameter optimization, time-series cross-validation, and uncertainty quantification using the conformal prediction (CP) method. The results show that the Transformer model achieves higher accuracy than the other models, with the lowest MAE, RMSE, and MAPE, and the highest R² and PICP. The Transformer model is effective at capturing long-term dependencies and providing a reliable forecast range, making it suitable for dynamic power systems.