Reservoir Uncertainty Evaluation - A Producing Gas Field Case Study
MetadataShow full item record
Once a field starts to produce, new information becomes available in terms of production data and measurements. By integrating this information to a reservoir model through a history matching process, an updated uncertainty evaluation of the reservoir can be formed. In the history matching process the current reservoir simulation model will be calibrated to match the new information. This is done to be able to represent the true dynamic behavior in the reservoir, hence get reliable predictions of the future reservoir behavior. The history matching problem is a non-unique problem, meaning that it has multiple solutions. This thesis presents a case study where history matching is applied to a producing gas field, through an assisted history matching process by the use of MEPO software (Schlumberger). The aim of the study is to get an updated uncertainty evaluation of the reservoir, three years after production start. The thesis will investigate the key uncertainties associated with the dynamic reservoir behavior in the field of study. To be able to capture and mitigate the uncertainties related to reservoir parameters, the ensemble based method Markov chain Monte Carlo is used with Bayesian updating. Utilizing the Bayesian framework imply to take a probabilistic approach to uncertainty. The probabilistic theory and its application to mitigate reservoir uncertainty through a history matching process, are reviewed and further applied in the thesis. In the study, the gas volume in place and the influence from the aquifer, in addition to the internal communication in the reservoir are considered the key uncertainties. The reservoir is divided into regions to be able to capture the uncertainty in different areas. Pore volume- and permeability multipliers for the regions are chosen as uncertain parameters in the study. The pressure in the field are matched by letting the algorithm sample the parameter values from the prior distributions for the uncertain parameters. The sensitivity simulations show the parameter influence on the bottom hole pressures in the wells and on the gas volume. The results show that the pore volume multipliers are the most influential parameters on the match. History matching is carried out in three scenarios. The ensemble of the sufficient history matches makes up the posterior distributions of the uncertain parameters. By analyzing them, it is found; presence of a possible barrier, influence from a small aquifer and a dominating parameter in terms of a pore volume multiplier. Further, the lack of convergence in the Markov chain is tested, the results did not imply lack of convergence in the sufficient scenarios, hence predictions are simulated. The field gas in place for the best matches are found to be in the range between 6.5 and 8 GSm3. The predictions show a total production period of 8-9 years before abandonment pressure is reached. Even though it is important to obtain a reasonable history match, it is at least just as important with a geologically truthful simulation model to be able to obtain reliable predictions. This is why a geologist s evaluation is essential to include in the study. Discussions with a geologist lead to geological interpretations from the results in the study that should be implemented and tested in a further study. The recommendations for future work are; (1) implement the suggested faults, (2) divide the dominating region into multiple regions, and further, depending on the previous results, (3) reduce the net-to-gross parameter values in the initial model.