Machine learning for predicting methane production and optimizing parameter in anaerobic digestion process

Abstract

Anaerobic digestion methane production is an important chemical means of waste recycling. Although the process route has been mature, there are still some difficultie, such as complex construction process, many variables, and difficult to find the influence law of the process parameters. Because traditional anaerobic digestion modeling is complex, the production volume can be predicted just and the influence mechanism of raw material characteristics and process parameters of the product is not clear. Accurate prediction of methane production and optimization of reaction process parameters are essential for the understanding the reaction mechanisms and optimizing the process parameters of anaerobic digestion process. Using Machine learning can skip the tedious process and select key features directly to predict methane production. In this work, First, model ADM1 as a relatively accurate anaerobic digestion prediction model, it provides data set for machine learning, addreses data quality and data quantity in anaerobic digestion process. Then,different machine learning algorithms were used to predict methane production and optimize process parameters. Finally, the most suitable machine learning algorithm for predicting the anaerobic digestion process was found. Using LightGBM and BPNN both can achieve more than 98% accuracy in predicting methane production and optimizing process parameters,. shortening the reaction time step, more accurate results can be obtained for LightGBM and BPNN. This work is a successful application of machine learning to anaerobic digestion processes. Applying AI can solve the common problems of energy production process and provide new ideas for the development of smart energy and smart engineering.