In sit characterization of the transport and sorption characteristics of coal seams from pressure transient data: An artificial neural network approach

This paper identifies the need for the development of specialized inverse solution techniques for the analysis of pressure transient data for the characterization of the transport and sorption properties of coal seams. Central to the proposed inverse solution technique is a specially structured artificial neural network (ANN). The most important transport and sorption characteristics of coal seams are butt and face cleat permeabilities, macropore porosity, Langmuir volume constant, Langmuir pressure constant, and sorption time constant. Conventional methods, such as type curve matching techniques and laboratory-derived measurements are handicapped in terms of being representative over the entire domain of interest. In this paper, we utilize analytical forward solution techniques for various combinations of characteristics of coal seams to generate training and testing patterns. These training patterns are shown to an ANN structured to learn the ubiquitous non-linear relationships that exist between coal seam properties, input conditions and the pressure transients measured. The trained networks are then extensively tested to ensure that they have effectively learned the training patterns. In the final stage of the development, the networks are run in the prediction mode. The responses of the networks are found to be in agreement with the input data. The developed networks are able to generate results generally within 5% error margin for permeability and porosity and 20% error margin for the sorption characteristics of coal seams. A number of examples highlighting the capabilities and degree of effectiveness of the network in characterizing the coal seams areincluded. Keywords: artificial neural networks (ANNs); coal seam characteristics; pressure transient analysis