Application of ANN-based Prediction Insights into the Stable Isotope Geochemistry of the Clean Energy Coalbed Gas

Abstract

The coalbed gas content, gas composition and stable isotope geochemistry of coalbed gas were investigated to study the generation mechanism of coalbed gas. The dry gas content of coalbed gas samples ranges from 2-10 m³/t. The composition of coalbed gas reveals that it consists of methane (∼52% to ∼99%) with higher hydrocarbon (0% to ∼12%) and traces of CO₂. The manuscript is designed to recognise the stable isotopes of coalbed gas samples using machine learning approaches and then provide a critical review of the gas origin from Jharia Coalfield. The artificial neural network (ANN) was constructed for this purpose, consisting of six parameters as the input and three as the model's output. The stable isotopes of coalbed gas samples from Jharia Coalfield were predicted by the model with the following ranges: -59.86‰≤ $$ ≤-19.31‰, -19.93‰≤ $$ ≤-7.95‰, while -275.74‰≤ $$ ≤-138.64‰. The wide range of isotope data imitates the complicated generation mechanism of coalbed gas; both thermogenic and biogenic methane are present in the coalbed gas. The carbon dioxide and methane index (CDMI), hydrocarbon index (HI), and gas dryness index (DI) were determined for the description of the Bernard and CD diagram (carbon-hydrogen diagram or Whiticar-style plot) to infer the origin of coalbed gas from the Jharia Coalfield. The experiment of the stable isotope analysis of the coalbed gas was also performed, which augmented the ANN-based prediction of stable isotopes by providing a strong correlation (R²>0.99). Van Krevelen’s diagram concludes that the coal samples fall in the window of type III gas origin. This research provides fundamental insights into the generation mechanism of clean energy coalbed gas.