Predicting Natural Gas Heating Value using Supervised Machine Learning Models*

Abstract

Heating value (HHV) is an essential parameter for evaluating natural gas quality. The existence of supervised machine learning models is therefore necessary for HHV prediction to ensure the eco-friendly and efficient utilisation of natural gas. This study aims to develop machine learning models based on the gas composition to accurately predict the HHV of natural gas. Three predictive models namely; decision tree, AdaBoost, and XGBoost models were used in the evaluation. Data samples from Jubilee, TEN, and SGN fields in Ghana were used. The study considered 721 data sets and the performance of each model was evaluated using R², RMSE, and MAE. Results obtained highlighted XGBoost model performs better than the other models. This was backed with an R² value of 92.9 % and RMSE and MAE error values of 2.002 and 1.195 respectively.