Recently, the Innovation Team of Clean Conversion and High-value Utilization of Farming Waste from the Institute of Environment and Sustainable Development in Agriculture, CAAS, accurately predicted the characteristics of digestion products in wet biogas projects via machine learning algorithms. The related findings have been published in   Chemical Engineering Journal .

Prediction of characteristics of products in wet biogas projects using machine learning algorithms

Anaerobic digestion is an important pathway for treating agricultural waste. In order to avoid deficiencies such as low accuracy and poor linear correlation in the prediction of digestion performance using traditional machine learning models, typical non-time series models (GBR and RF) and time-series models (LSTM, CNN-LSTM, and DA-LSTM) after hyperparameter optimization were chosen in this study to predict the characteristics of digestion products in wet biogas projects.

The ideal GBR model for methane content prediction was obtained, and DA-LSTM was superior to LSTM and CNN-LSTM for the prediction of biogas production. The study established reliable machine learning algorithms that can provide important theoretical and technical support for process regulation and efficient biogas production in wet biogas projects. Meanwhile, the methodological framework involved, including input data preprocessing, model training and optimization, accuracy validation and generalization capability enhancement, can provide methodological guidance for model prediction in other renewable energy projects.

This study was supported by the China Agriculture Research System of MOF and MARA, National Natural Science Foundation of China (32301727), and Agricultural Science and Technology Innovation Program (ASTIP) of China.

Linkage: https://www.sciencedirect.com/science/article/pii/S1385894724070736