Assessing machine learning tools for methane emission prediction from POME treatment in Malaysia

Abstract

Palm oil mill effluent (POME) treatment is an anthropogenic activity contributing to global warming through methane emission. The inability to address this issue would deem true the catastrophic impacts of global temperatures exceeding 2 °C as was predicted by the Intergovernmental Panel on Climate Change (IPCC) in 2015. Little research and development exist on GHGs monitoring and methane emissions in POME treatment facilities as opposed to research on improving biogas production. A methane emission prediction tool based on machine learning models and tools can address this problem and consequently facilitate the development of efficient carbon neutrality approaches in POME treatment plants. In this study, six regression models were explored alongside their kernels using eight predictors, linking towards methane emission volume. The best model found was support vector machine (SVM), producing performance metrics for R2 and RMSE with values of 0.45 and 0.749, respectively. HIGHLIGHTS Selection and utilisation of a suitable machine learning algorithm for the prediction of CH4 emissions.; Obtaining raw data on POME treatment for database development and understanding the nature and quality of the dataset using available data from pre-processing methods.; Integrating the developed database into the CH4 emission tool.; Comparing factors influencing CH4 emissions.; Determining the highest influencing parameters.;

Keywords

• machine learning
• methane
• palm oil mill effluent (pome)
• prediction
• support vector machine (svm)
• transformation