Probabilistic Reserve Estimation Constrained by Limited Production Data: AnIntegrated Approach

摘要

Large uncertainties in structure and facies had been recognizedrnin a major gas field in Pakistan after early production. Thernconventional reserve estimation methods had failed inrnproviding a reliable estimate of gas-in-place (GIIP). It wasrnpossible to get a good history match of one-year productionrndata using a wide range of GIIP through a slight andrnacceptable adjustment of porosity and permeability. Thernresulting possible range of GIIP could easily vary by a factorrnof 1.5. Structural uncertainties did not warrant volumetricrnestimates either. Material balance technique was questionablerndue to non-uniform drainage of the reservoir. Clearly theserndeterministic techniques of reserve estimation were notrnapplicable at this stage of production considering therncomplexities of the reservoir. A probabilistic technique wasrntherefore developed that addressed both static and dynamicrnuncertainties in an integrated approach while honoring thernavailable production history. Combined treatment of static andrndynamic uncertainties also ensured a better coverage of thernentire sample space, thus making the probabilistic approachrnmore reliable.rnLatin Hypercube Sampling (LHS) helped minimizing thernnumber of simulation runs while providing a reasonablerncoverage of the sample space. Yet we ended up with almostrn1500 simulation runs. The process of history matching,rnranking and keeping track of all these simulation runsrndemanded an innovative workflow. A number of softwarerntools were used to automate and optimize this process. Out ofrn1500 simulation runs, the 200 best runs having minimumrnobjective function through history matching were selected.rnThese runs were later used for production forecasting, forrnproviding a range of reserves, and for sensitivity analysis tornidentify the most influential variables. Structure and NTG were identified as the two most critical variables for *GIIPrnwhile residual gas saturation was identified as an additionalrnsensitive variable for reserves. Different geostatisticalrnrealizations had little impact on GIIP or reserves.rnDeterministic approach had resulted in GIIP from 1 to 1.7rnReservoir Volume Units (RVU)*. Probabilistic estimates ofrnGIIP, in comparison, ranged from 1.2 to 1.6 RVU. Morernimportant than a reduction in the range of reserves was the factrnthat this approach had considered all major uncertainties, staticrnand dynamic, before estimating GIIP and reserves. At thernsame time, these reserves were in harmony with the actualrnfield performance so far. This makes these numbers morernreliable and the probabilities like P10 and P90 morernmeaningful.