For reservoir characterization, the subsurface heterogeneity needs to be qualified in which the distribution of lithologies is an essential part since it determines the location and migration paths of hydrocarbons. Preliminary analysis of well-log data could help to identify various lithologies in a one-dimensional direction (depth), while the lateral information is missing because of the sparse locations. On the other hand, a larger areal coverage of the target reservoir could be provided by seismic data, and from the inversion thereof, inferences of lithologies could be made. However, just like other geophysical inversions, translation of seismic inversion results to these categorical variables (lithologies) is a non-unique problem, which means that different lithologies could produce the same, or similar, property responses. In order to mitigate this problem, geological prior information should be introduced in the sense of Bayes’ theorem. Thus, the main motivation for this thesis is to investigate the usage of geological prior information in the classification of reservoir lithologies from properties obtained from seismic inversion. Different methods have been tried in this process in order to fully understand their performances and to make comparisons.
|Award date||8 Dec 2017|
|Publication status||Published - 2017|
- seismic inversion
- Markov processes
- reservoir lithology