Forecasting peak demand for electricity using time series models and artificial intelligence models

Accurate power demand forecasting is essential for maximizing the efficiency of power grid operations.In this paper, daily peak electricity demands are predicted by using time-series models and artificial intelligence (machine learning and deep learning) models and their performances are compared.The analysis reveals that the XGBoost method outperforms others for one-step-ahead forecasts, while deep learning models, LSTM and GRU, exhibit superior performance for forecasts from two to seven steps ahead.For time-series models, predictions based on bias-corrected log-transformation and back-transformation methods yield better results than those using raw data.