Optimization and Prediction of Daily Methane Production for Food Waste By using The Artificial Neural Network (ANN)

This study aimed to know how a neural network model can be used to predict the behaviour of particular methane production using a three-layer (3:12:1) feed forward back-propagation algorithm and the logsig-purelin transfer function to maximize maximum methane prediction. The artificial neural network performs admirably compared to the daily methane generation of (vegetable waste 5%, lipids 12.5%, meat residues 12.5%, cow dung 10% and water 60%) at various temperatures (19–30 °C) for 64 days. Among the three input variables (feedstock, temperature, and time), feedstock has the strongest correlation with the particular methane production value. Thus, the research validates the ANN model’s ability to anticipate the biogas production curve’s behaviour and forecast the optimal substrate temperature for maximum biogas production. The regression value(R) for the digester was 0.95769. The capability of ANN modelling acts as a preliminary draft and significantly lowers the time required for methane generation in-line control. The daily methane production was predicted at three temperatures (20, 25, and 30 oC) for 75 days, and it found out the best temperature reaction was 30 oC.