Experimental Design Based History Matching of Reservoir Simulation Models for Champion Field

Abstract

This reference is for an abstract only. A full paper was not submitted for this conference. Abstract The Champion Field is located 40 km offshore Brunei, northern Borneo. It is a mature field with over 250 producing wells, oil production commenced in 1972, and the current recovery factor remains below 20%. The field comprises about 20 NE-SW oriented elongate fault blocks separated by East hading extensional faults, and there are about 100 individual reservoir units, giving a total of more than 500 hydrocarbon bearing compartments. To date, 37 water injectors have to varying degrees stemmed a part of the production decline. As such, redevelopment of this giant field calls for a large investment in water injection. How do you decide to make such a large investment when faced with such a complex field, where credible production and injection data is limited? A 60 man year effort lasting 4 years has analyzed the structure, the geology, and other properties including production and pressure data and all of the uncertainties associated with all these factors. Faced with mountains of data, and its associated large uncertainty ranges the team employed experimental design in the history match and forecast redevelopment workflow. Relatively large complex static and dynamic models (1,000,000+ active cells with complex fault systems) were used for the history matching exercise with more than 200+ history matching responses from over 30 years production and injection history. In the course of integrated modelling, subsurface uncertainties were evaluated and reviewed intensively and the ranges of uncertainties of the subsurface properties were defined independently. Experimental Design based history matching workflow has been applied to this large and complex simulation model with long history. The subsurface uncertainty factors were evaluated their impacts to the important history matching responses and screened to select the high influential factors. It is also important to make sure that the combination of the original uncertainty ranges (solution space) should covers target solutions (history matching objectives). Appropriate numbers of simulation runs were designed using combinations of the screened uncertainty factors. The response surface models (RSM) for the history matching responses have been generated using these simulation results. The uncertainty ranges of the selected factors were successfully conditioned by applying the error band to the historical data with RSMs. After the conditioning, the new uncertainty ranges were tested with the simulation runs to confirm the expected range of outcomes through blind tests. Conditioned models were then used in the production forecasts with re-introduction of low impact discarded factors and the introduction of new development related factors. Workflows established in this study were successfully applied to handle the large amount of data and speed up the history matching process while still handling all of the important factor uncertainties in order to quantify the impact of these factors on specific reservoir performance responses. This paper describes a direct approach that allows engineers and geoscientists to split out important factors associated with an exploitation process via experimental design history matching and then, to proceed with these factors into the forecast process to capture the full range of oil development outcomes. The process goes further to highlight which factor uncertainties need to be better understood before project commitment and gives the geoscientists and engineers a better method to address all the important factors and uncertainties to arrive at a better decision.