Key wastes selection and prediction improvement for biogas production through hybrid machine learning methods

As history shows, a linear production process results in a waste of resources. Using renewable energy can not only reduce waste but also protect the earth's resources. However, due to the high input costs of biogas manufacturing and a lack of standard manufacturing procedures, there have been few predictive models generated using multiple algorithms to analyze key input wastes and to predict output in the field of biogas generation. The lack of key wastes analysis may result in reductions in both the optimization of production processes and the stability of production, eventually affecting the popularity of biogas generation. Therefore, this research proposes a hybrid machine learning method based on random forest (RF) and long short-term memory (LSTM) analysis for redefining the key factors and improving the prediction of biogas generation output. The analysis adopted input and production output data from an existing biogas plant. Regarding the predication for biogas production, the performance of the proposed hybrid model exhibits 20% more accuracy than its nearest competitor among traditional analytic models. Coupled with the assistance of a hybrid machine learning method, the identification of key factors can be improved, providing users with suggestions and guidance for modifying parameters. Through a precise control of final output and manufacturing process, this methodology may be expected to help reduce costs and accelerate the popularity of biogas.