Underground coal methane gas forecasting using multivariate time series with one and two auxiliary variables

Prediction of methane gas concentrations in underground coal mines has been a topic of interest for academia and the mining industry for decades as methane gas is considered a significant hazard to mining safety and productivity. This study utilizes multivariate time series forecasting approaches to predict methane emissions in an active mine. The multivariate method incorporates the past values of methane gas concentrations as a predictor of its future values but also integrates the correlations between gas concentration and barometric pressure as well as barometric pressure and coal production rate in predictive models of methane gas emissions. The data used in this study were collected from the atmospheric monitoring systems of an underground coal mine operation in the eastern USA. The data were stored, pre-processed, and statistically analyzed to develop multivariate forecasting models based on the Vector Autoregression (VAR) and Autoregressive Integrated Moving Average with Exogenous Input (ARIMAX) methods. The optimal model for each method was selected based on the Akaike Information Criterion. The forecasting performance and accuracy of each model were assessed using cross-validation to establish the best overall model. It was concluded that both VAR and ARIMAX could, in most cases, adequately forecast methane gas concentrations in underground coal mines.