The reliability of Pareto-based history matching decreases as the number of objective functions increases because the probability that trade-off solutions exist in the non-optimal domain increases exponentially. This study developed an evolutionary algorithm to overcome inefficiency of the objective constraint by introducing preference-ordering and successive objective reduction to the conventional multi-objective optimization module. The former enhances the convergence speed towards the Pareto-optimal front by pruning any unqualified geomodels, and the latter improves the optimization efficiency by excluding redundant production data from the fitness evaluation. This integrated model consistently reduced the data mismatch between the observed and calculated production, so that overcame the divergence problem in multi-objective optimization as well as the scale-dependency problem in single-objective optimization. This calibration process predicted the future performance better than the typical optimization schemes from equiprobable geomodels, preserving the diversity of feasible solutions, thereby assessing uncertainties in production forecasts for both the field and wells.
- Evolutionary algorithm
- Pareto-based history matching