Optimisation and economic analysis of industrial-scale anaerobic co-digestion (ACoD) of palm oil mill effluent (POME) and decanter cake (DC) using machine learning models: A comparative study of Gradient Boosting Machines (GBM), K-nearest neighbours (KNN), and random forest (RF)

Abstract

The anaerobic co-digestion (ACoD) of palm oil mill effluent (POME) with decanter cake (DC), has been proposed as a solution to enhance biogas production in industrial-scale anaerobic covered lagoon systems. Despite the potential of ACoD, there is a limited exploration of machine learning techniques, such as Gradient Boosting Machines (GBM), K-nearest neighbours (KNN), and Random Forest (RF), specifically in the context of industrial-scale facilities and anaerobic co-digestion. In this study, these machine learning techniques were employed to develop prediction models for biogas production and BOD removal using three months of operational data from a local biogas plant, with GBM yielding the highest prediction accuracy (R2 = 0.9940, MAE = 0.0101, RMSE = 0.1438). Subsequently, GBM was used in the optimisation process to determine the optimal conditions for temperature, pH, and dilution ratio of DC in the anaerobic digester. The optimal conditions were found to be 40 °C, 7.1, and 0.11, respectively, representing a 21.58 % and 26.03 % increase in biogas production and BOD removal. This integrated approach provides a sustainable solution for maintaining year-round biogas production even when the supply of POME is limited, thereby contributing to renewable energy generation and waste management in the palm oil industry.