Predictive Modeling of Biogas Production Using Machine Learning

Abstract: This project focuses on predictive modeling of biogas production using machine learning to improve efficiency, reliability, and scalability. The dataset, sourced from Professor Jackson Milano’s research at Universidade Positivo, spans 14 months of continuous monitoring on a ranch in southern Brazil, using cow manure in four biodigester configurations. The data was cleaned and preprocessed, and machine learning algorithms such as Linear Regression, XGBoost, LightGBM, Random Forest, and TensorFlow were used to develop models for estimating biogas yield. The iterative training process ensured high predictive accuracy. A user-friendly web interface was developed to allow real-time interaction with the model, enabling users to input parameters and receive biogas output predictions. This project showcases the potential of machine learning in optimizing renewable energy systems, promoting sustainability, and smarter energy management.