Energy recovery from municipal waste using machine learning algorithm to produce biogas

The application of machine learning in the field of renewable energy witnessed progressive growth in recent years. Historically, its role in improving efficiency in energy systems, optimization of demand and production and renewable generation and/or price forecasting, has been well established in the literature. This paper aims to apply the AI approach in the field of energy recovery from municipal waste collection to produce biogas. Generally, the correlation between the collective waste and the amount of energy that can be extracted is not straightforward. Several aerobic and anaerobic digestion processes are performed inside the collection. On the other hand, climate condition and the quality of the ground itself has a determining role to accelerate or limit the chemical transformation. As a result, the available models in this field are mostly fit to real data obtained from one specific case study that cannot be applied to other cases. The methodology conducted in this paper contains municipal waste database in Italy from 2017 to 2022 that is used to train the Machine learning models. After creating the models, the new sets of data are imported to verify model accuracy. In this step, six algorithms are introduced: simple tree, linear regression, support vector regression, gaussian process, kernel and artificial neural networks. The final results are compared with the available models in this field. Three-step optimization is applied to improve the accuracy of the machine learning model including manual filtration, principal component analysis and Auto minimal dependency detection. Eventually, the final results show accuracy in the range of 92-99 % depending on the used algorithm. On the other hand, using the available mathematical mode with the same input, the maximum accuracy of 90% is achieved. The simulation is carried out using MATLAB ML toolbox.