A hybrid statistical and machine learning based forecasting framework for the energy sector

Energy prices are sensitive and volatile due to supply-demand imbalances, policy changes, new environmental targets, and technological breakthroughs. Energy security, sustainability, and affordability are key to policy design and societal advancement, making accurate energy price forecasting essential. Despite the existence of various methods for forecasting energy prices, no single method consistently outperforms the others. In this respect, the Energy Price Index (EPIC) framework is presented with extended forecasting capabilities for 56 different energy products using 33 unique time series. Statistical and machine learning forecasting methods of different natures have been incorporated into the framework, enabling the forecasting of energy prices up to 14 months ahead. Each energy product's historical prices are analyzed for seasonality, trends, and outliers. The results reveal the superiority of deep neural networks over statistical methods, while at the same time highlighting the significant importance of proper tuning of the hyperparameters of a neural network.