This paper presents a case study using a Multipurpose Environment for Parallel Optimization with application to assisted History Matching. Evolutionary Algorithms and deterministic optimization schemes are integrated into a workflow controlling a large number of parallel reservoir simulations.
The practical applicability of a software assisted history matching process is illustrated on basis of a real case study. The history matching process is discussed for an inverted 5-spot water flood pattern that had been subject to an extensive program of reservoir pressure and water saturation monitoring. Results are analyzed and compared to traditional history matching to identify the potential added valued and increased efficiency.
The quality of the manual history match was reproduced for pressure data and significantly improved for saturation data. The resulting model has proven to show better forward modeling characteristics defined by matching of some blind test data. Due to highly constraining boundary conditions defined by pressure and saturation data, it was not possible to generate multiple solutions to this specific problem. Within the given constraints, the model tuning parameters had to be close to the values of the reference case to be able to match such constraining pressure and saturation data.
In conclusion, the software assisted history matching process was able to improve the quality of the match within less time and effort. A lesson learned process is discussed focusing on the engineer acquiring more information and improving the understanding on reservoir uncertainties and the reservoir model behavior.
Geological models derived from static data, such as geological, well log, core and seismic data, often fail to reproduce the reservoir production history. History matching is defined as the process of reconciling geologic models to the dynamic response of the reservoir. The main purpose of history matching is building a numerical simulation model which is consistent with the entire available reservoir data, i.e. geological, petrophysical and SCAL data as well as production data including field and well pressure, flow rates, water cuts and gas oil ratios.
In order to obtain an acceptable description of the reservoir by history matching, many different simulation runs in completely different regions of the search space must be performed. In order to capture reservoir model uncertainties within the range of model parameter uncertainties, a variety of models should be generated. They will not be distinguishable with respect to the reproduction of history data but may deliver different predictions of future reservoir performance.
Manual history matching process often tunes limited set of parameters to reach to only one acceptable non-unique history matched model. In addition to this serious limitation, manual history matching is time consuming, and faces a real challenge in terms of keeping track of the model response to parameters changes and their combined effect on the field, sub reservoir and wells levels.
Some known giant fields in the world are carbonate reservoirs characterized by a high level of heterogeneity. This level of heterogeneity necessitates the understanding of reservoir uncertainties in all levels of the reservoir modeling process including: data acquisition, geological static modeling and dynamic simulation modeling. Manual history matching of the simulation of these kinds of reservoirs often fails to cover important reservoir uncertainties. Due to the availability of increasing computing resources, software assisted optimization techniques with application to reservoir simulation are given more and more attention. With these techniques, turnaround times for creating new models and updating old models has been significantly reduced.