Putting Error Bars on Reservoir Engineering Forecasts

Abstract

Summary. Uncertainties about reservoir characteristics[hydrocarbons originally in place (HCOIP), faults, sand/shalecontinuity, aquifer size, relative permeability, etc.] and aboutoperational factors (regularity, pump lifetimes, etc.) oftencombine to yield a significant composite uncertainty aboutproduction forecasts. This paper presents a methodology for production forecasts. This paper presents a methodology for converting reservoir. geological, and other uncertainties intoproduction forecast and reserves uncertainties. The engineer's job production forecast and reserves uncertainties. The engineer's job is not complete until a base-case forecast can be seen inconjunction with plausible upside and downside forecasts. Theoverall field development strategy and topside facilities andcapacities can then be chosen to prepare as much as possible forthe error bar on forecasted rates at a given point in a field'slife (which may be expensive). Error bars also can be used to judgethe risk of the project and may recommend additional datagathering, long-term testing, pilots, or a phased (invest as youinvestigate) development. Only when the individual componentuncertainties and the total uncertainty level are uncovered andtheir implications have been analyzed does a sufficient basis formaking an informed decision exist. Introduction Historically, most reported postanalyses of field developmentshave revealed lower production rates and smaller reserves thanpreproduction forecasts indicated. Preproduction uncertainties preproduction forecasts indicated. Preproduction uncertainties about reservoir characteristics, inadequate representation and lackof realistic geology in numerical models and unforeseen operationaldifficulties and constraints (topside misfits) may have been themain causes of erroneous (optimistic and pessimistic) forecasts. Specialists within each discipline should question theaccuracy of their input data in the process of quantifying theuncertainties of the components they introduce. The challengefacing forecasters in the 1990's is to uncover and combine themajor relevant component uncertainties into a base-case forecastwith associated error bars. Today, few professionals believe that oil price increaseswill "save" (as they did during 1974-83) oil and gas field developmentsbased on overly optimistic forecasts of future oil and gasproduction rates. Smaller fields, geologically more complex production rates. Smaller fields, geologically more complex fields, smaller economic margins, and less robust projects (e.g., EOR and infill drilling) are typical petroleum engineering issuesin many hydrocarbon-producing regions of the world in the 1990's. For this reason, oil companies need a systematic method forquantifying the composite technical uncertainty (in productionrates and reserves) and the compounded economic risk [in netpresent value (NPV) and other economic indicators] associated with present value (NPV) and other economic indicators] associated with field developments and incremental projects. This paper presents a possible method for convertingreservoir, geological, and other uncertainties into productionforecast and reserves uncertainties. The inclusion of technicaluncertainties into project economic-uncertainty evaluations also isdiscussed. The suggested approach is based on sensitivity analyses(numerical simulation of plausible states) followed by statisticalsimulation (which accounts for the probabilities of the variousplausible states). plausible states). Quantifying Uncertainty Practical experience and sensitivity analysis allow Practical experience and sensitivity analysis allow identification of which parameters are the most critical in termsof their influence on the amount, position, accessibility, in-situflow, and long-term production of hydrocarbons from subsurfacereservoirs. The influence of the identified parameters on theresult (e.g., the HCOIP and the production profiles) may bequantified by running one-parameter-at-a-time sensitivity studies(OPAATSS) on the models that convert input into output(HCOIP-expression. numerical simulation model). OPAATSS does notalways work well in problems with interactions (simultaneousparameter variations), and a more complex experimental design for parameter variations), and a more complex experimental design for the sensitivity study is often needed to capture nonobviousparameter interactions. Another general limitation with parameter interactions. Another general limitation with sensitivity analysis is that one only looks at the consequence ofchanging input values; one does not assess the likelihood of suchvalues and, hence, how likely the various outcomes are. Genrichand Sommer presented a novel approach to sensitivity analysis thatalso results in influence ranking of input parameters. Repeated simulations (in a statistical sense) are frequentlyused to quantify composite uncertainty caused by individualparameter uncertainty. The output from the statistical simulations parameter uncertainty. The output from the statistical simulations is an outcome histogram or a probability density function thatgives a quantitative measure of the output range (and theprobability of taking on values within this range). probability of taking on values within this range). The approach suggested here is a stepwise procedure based onsensitivity analysis, subjective assessment of probabilities, andstatistical simulation for an assessment of the compositeuncertainty. A large number of publications deal with the quantificationof uncertainty and risks in exploration, production geology, reservoir engineering, economics, and many other disciplines. Forspace reasons, only a statistical sample of these publications isprovided. provided. Several techniques exist to obtain the distribution of an outputvariable, given the known or assumed statistical distributions ofthe inputs. JPT P. 732