Novel Reservoir-Simulation Methods Enable Direct Optimization of Expansion Decisions in Giant Heavy-Oil Field

Abstract

Giant, geologically complex heavy oil fields can take decades to develop, sodevelopment decisions made early in the life of the field can have long-rangeimplications. Decision and risk analysis (D&RA) is often used to makedecisions that will maximize risk-adjusted economic benefit. Unfortunatelyin large heavy oil fields, D&RA can be very challenging due to the largenumber of variables and the endless number of development and expansionscenarios to analyze. The time needed to complete a D&RA can becomeprohibitive when full-field reservoir simulation is the main tool forforecasting primary production and well count, with one simulation taking manyhours or days to complete. This paper describes two new simulation tools developed to overcome thesechallenges: 1) a method for populating a model with hundreds-to-thousands ofhorizontal wells, and 2) a method to quickly and directly optimize expansiondecisions. A semi-automated spreadsheet-and-simulation method was developed to quicklyplace and select hundreds-to-thousands of hypothetical/future horizontal wellsin a multi-million grid- block model. Because the method automaticallyaccounted for all model static properties and their effects on dynamicproduction response, the hypothetical wells had productivity characteristicsvery similar to actual drilled wells placed in the model. A multi-variant non-linear interpolation method was developed that enabledfull-field forecasts - for any combination of acreage allocation, well count, drilling order, and expansion rate constraint - to be calculated in less than 5seconds, compared to about 20 hours for traditional simulation. Extensivevalidation work showed that well count and production curves from thespreadsheet virtually overlaid those obtained using traditional simulation ofthe particular expansion scenario. Such close agreement was possiblebecause the basis of the spreadsheet forecast was utilization of traditionalsimulation forecasts from a handful of relevant cases. A key breakthrough beyond just fast forecasting was the coupling of thefollowing three components inside the same spreadsheet: the fast forecastingmethod, calculation of an economic indicator/objective function (NPV), andcommercial optimization tools. This linkage made possible, perhaps for thefirst time (at least at this scale), realization of direct optimization of anydevelopment scenario in a matter of minutes to a few hours, depending on thenumber of variables being optimized. Introduction The field of interest was areally very large and vertically geologicallycomplex, with multiple stacked reservoirs. Rock and fluid propertiesvaried significantly with reservoir, depth, and areal location. The crudewas very heavy, and an expensive processing facility was required for transportand marketing. The oil in place was large enough that the reserves werelimited primarily by the capacity of the processing facility (Fig. 1) and theduration of the production agreement (assuming a sufficient number of economicwells could be drilled to fill the facility). Early in the field life, there was the possibility of expanding theoperation by building a second facility. One of the critical decisions wasthe size of the second facility. Because it was early in the life of thefield, there was little production history and the field was not fullydelineated. Therefore, there remained significant uncertainty in reservoirperformance and thus in the number of wells required to fill a secondfacility. The situation certainly called for D&RA, but the decision tree wascomplex, comprising over 100 branches. The traditional solution would havebeen to run multiple full-field simulation cases to attempt optimization ofeach branch of the decision tree. However, that approach would have beennearly impossible to complete, because each simulation took a day to run (evenusing parallel processing and a dozen cpu's) and optimization would have been atrial-and-error process.

Bridging the Gap Between Deterministic and Probabilistic Uncertainty Quantification Using Advanced Proxy Based Methods

“…(Cottini-Loureiro & Araujo, 2005) (Cullick, Heath, Narayanan, April, & Kelly, 2004) (Cullick, Narayanan, & Gorell, 2005) Onwunalu, & Aziz, 2006) (Bailey, Couet, & Wilkinson, 2005) (Osterloh & Menard, 2007) (Ozdogan, Sahni, Yeten, Guyaguler, & Chen, 2005) (Shelkov, et al) (Baris & Horne, 2001) (Guyaguler & Horne, 2001). A proxy based approach may be found in (Jutila & Goodwin, 2006).…”