Enhancing Biogas Energy Production Forecasting with Box-Cox Transformation and Prediction Interval Methods

Biogas, a renewable energy source, plays a crucial role in reducing greenhouse gas emissions and promoting sustainable energy production. Its dual benefits of waste management and energy generation make biogas a key component of renewable energy strategies. However, forecasting biogas energy production remains challenging due to inherent variability and uncertainties in feedstock composition and co-digestion processes. This study addresses these challenges by comparing traditional point predictions with uncertainty-aware approaches. Initially, a simple neural network was applied to the original data, achieving good accuracy for point predictions but failing to account for production uncertainty, limiting its practical applicability. Recognizing the importance of capturing variability, this research integrates a hybrid approach utilizing the Box-Cox transformation and the Lower Upper Bound Estimation (LUBE) method. The Box-Cox transformation normalized the skewed data, improving the model's ability to capture complex variations and refine Prediction Interval Normalized Average Width (PINAW). The LUBE method, applied after transformation, provided prediction intervals that were both narrower and more reliable than traditional methods. These enhanced intervals significantly improve decision-making by providing better insights into production variability. Policymakers and investors can leverage these tools for strategic planning and resource allocation, aligning with sustainable development goals (SDG-7) by fostering confidence in renewable energy production from biogas.