The Application of Neural Networks to Underground Methane Prediction

The prediction of underground methane emission from longwall coal faces has been extensively researched for a number of years and numerous prediction methods have been devised. Many coal mines employ sophisticated environmental monitoring equipment which generates large amounts of data including, methane concentration, air velocity and barometric pressure. Presently this information is mainly used only to indicate alarm and danger levels of the monitored data for management purposes, however, the potential uses of this data are enormous[1]. This information could be used to good effect within a production/methane emission strategy as detailed in a previous paper[2]. The paper presents a practical application of artificial neural networks to underground methane prediction. Data on methane concentration and coal production has been obtained from an underground coal mine. The data is used in a neural network architecture and forecasts of methane concentration have been obtained. The forecasts are examined and there is a discussion of the potential use of the predictions from neural networks in the field of mining production control.