Maunde A, Alalade TF, Raji AS and Haruna IV
Petroleum reservoirs are characterised by various feature which are indicative of the geologic processes that resulted in the formation of such rock bodies. These geologic processes are responsible for the major uncertainties surrounding oil and gas production. There is therefore the need to quantify these uncertainties to an acceptable level to allow for the certain risk evaluations on petroleum projects. The process of history matching involves creating of reservoir models which mimic the observed reservoir performance to some acceptable extent. This paper evaluates two numerical schemes by which the history matching process is optimised with the subsequent utilisation of the models obtained from this process to infer certain properties deemed to be representative of the reservoir future performance. Single objective and multi objective particle swarm optimisation algorithms are used in optimised history matching of the synthetic PUNQ-S3 reservoir with the results from the two schemes put forward to a Bayesian evaluator for forecasting. The results obtained suggest the multi objective particle optimisation scheme not only produces good quality history matches but it also converges faster. With regards to forecasting, the models obtained from both schemes did not reflect the observed well bottom hole pressures. However, the multi objective scheme provided better forecasts of the field total oil production relative to the single objective scheme with the truth case being reflected. Moreover, the uncertainty intervals created from the multi objective scheme are wider than those generated from the single objective scheme.
Share this article