Effect of Grid Size in Risk Assessment of Petroleum Fields

摘要

It was shown recently that it is important to evaluate riskrnthrough probabilistic methodologies that involve a highrnnumber of simulation models because of the number ofrnuncertain attributes. Geological modeling yields reservoirrnmodels that are represented through fine grids with millions ofrnblocks. A probabilistic risk evaluation based on such gridsrnwould require a very high computational effort. Therefore, anrnupscaling procedure is necessary to reduce the grid size but itrnis difficult to select a grid size that could represent an adequaternbalance between precision of risk assessment andrncomputational effort.rnThe methodology applied to quantify risk involves arnsensitivity analysis in order to reduce the number of criticalrnattributes and the simulation of reservoir models obtainedrnthrough all possible combinations of these attributes. After thernsimulation of the models, a statistic treatment is used tornevaluate the risk involved in the process. Several proceduresrncan be used to speedup the process; however the number ofrnsimulation runs may be very high. Upscaling of the simulationrnmodels can decrease significantly the computational effort andrnglobal time of the risk analysis process but it can also yield anrninadequate risk assessment.rnIn this paper the effect of the grid size on the process isrnevaluated. It was developed a methodology (1) to select anrnadequate grid size and (2) to speed up the risk analysisrnprocess. The choice of geological representative models fromrncoarse grid risk evaluation can to be useful to represent thernrisk on fine model, avoiding the simulation of all fine models,rnyielding a significant speedup up of the process.rnPractical applications of upscaling in a probabilistic riskrnassessment, that use the concept of representative modelsrnselected to characterize geological uncertainties, are shownrnthrough calculations performed in a petroleum fieldrnrepresented by a fine grid simulation model with geologicalrnuncertainties.