Download files
Abstract
Coal seam gas (CSG) is gaining global interests due to its natural abundance and environmental benefits in comparison to more traditional energy sources. To efficiently recover this clean and abundant energy, a detailed understanding of the fundamental properties of coal is required, which makes the coal characterisation be of paramount importance. However, coal is highly heterogeneous at multiscales, such that the characterisation of coal lags behind that of conventional reservoir rocks. X-ray microcomputed tomography (micro-CT) has been widely applied to obtain the 3D digital representation of the coal samples. However, micro-CT images always suffer from a trade-off between the sample size to be scanned and the obtained resolution, and commonly pose a major challenge for direct numerical simulations due to the simulation instabilities. This thesis designs a complete and comprehensive coal characterisation framework, and demonstrates the capability of a digital model, named “Digital Coal”, can be used to characterise the multiscale structure of coal as well as petrophysical properties. The “Digital Coal” model developed in this thesis can preserve multiscale features of real coal samples, including the coal lithotype distribution on centimetre-length scale, cleat network and mineralisation on millimetre- to centimetre-length scale as well as cleat rough surface structure on micrometre-length scale. Our developed “Digital Coal” integrates multiscale geostatistics that are obtained from micro-CT data of different resolutions. Hence, it is not restricted by imaging resolution and can be constructed with extended domain size. Therefore, “Digital Coal” can act as an alternative to the segmented micro-CT images, such that we can avoid the challenges that are inherent to micro-CT images, such as the resolution limitation and segmentation errors. This study bridges coal properties at multiple scales, which can be potentially applied for the prediction of CSG and water production on reservoir scales by populating the model into reservoir simulators.