History matching of enhanced coal bed methane laboratory core flood tests

Enhanced coalbed methane (ECBM) involves the injection of a gas, such as nitrogen or carbon dioxide, into the coal reservoir to displace the methane present. Potentially this strategy can offer greater recovery of the coal seam methane and higher rates of recovery due to pressure maintenance of the reservoir. While reservoir simulation forms an important part of the planning and assessment of ECBM, a key question is the accuracy of existing approaches to characterising and representing the gas migration process. Laboratory core flooding allows the gas displacement process to be investigated on intact coal core samples under conditions analogous to those in the reservoir. In this paper a series of enhanced drainage core floods are presented and history matched using an established coal seam gas reservoir simulator, SIMED II. The core floods were performed at two pore pressures, 2 MPa and 10 MPa, and involve either nitrogen or flue gas (90% nitrogen and 10% CO₂) flooding of core samples initially saturated with methane. At the end of the nitrogen floods the core flood was reversed by flooding with methane to investigate the potential for hysteresis in the gas displacement process. Prior to the core flooding an independent characterisation programme was performed on the core sample where the adsorption isotherm, swelling with gas adsorption, cleat compressibility and geomechanical properties were measured. This information was used in the history matching of the core floods; the properties adjusted in the history matching were related to the affect of sorption strain on coal permeability and the transfer of gas between cleat and matrix. Excellent agreement was obtained between simulated and observed gas rates, breakthrough times and total mass balances for the nitrogen/methane floods. It was found that a triple porosity model improved the agreement with observed gas migration over the standard dual porosity Warren-Root model. The Connell, Lu and Pan hydrostatic permeability model was used in the simulations and improved history match results by representing the contrast between pore and bulk sorption strains for the 10 MPa cases but this effect was not apparent for the 2 MPa cases. There were significant differences between the simulations and observations for CO₂ flow rates and mass balances for the flue gas core floods. A possible explanation for these results could be that there may be inaccuracy in the representation of mixed gas adsorption using the extended Langmuir adsorption model.