A statistical analysis of monitored data for methane prediction
This research describes an investigation into the application of a statistical method for the prediction of methane concentration in longwall coal districts. An important and necessary part of the research was the acquiring of representative mine environmental and coal production data and a number of shortcomings were identified in this area. The monitored data was used to build univariate time series models of general air body methane concentration, air velocity, barometric pressure, coal production and methane drainage variables of varying timescales according to the Box-Jenkins method of time series analysis. The univariate models were used to identify causal relationships between methane concentration and its explanatory variables. Coal production was found to be the dominant variable in the determination of the quantity of methane emitted and where appropriate, multivariate time series models were built in which expressions for methane concentration in terms of coal production were obtained. Forecasts of methane concentration values were generated from both univariate and multivariate models and a comparison was made of their forecasting capabilities. Finally, suggestions were made as to the potential use of time series models for application to mining process control.