History matching and forecasting of geologically complex reservoirs are
challenging aspects for efficient reservoir management. The foremost reason is the high level of uncertainty that exists in the reservoir models because of the limited, sparse, and multi-scaled reservoir data available. The objective of the history matching is to adjust the model and its parameters (e.g., permeability and porosities) such that the simulation model is able to reproduce the ell-flo-rate and pressure histories reasonably. The result of the history-matching process is a ne simulation model that can be very different from the original geological model. Hoever, this history matching is an inverse problem, meaning that there is no uni!ue solution. "ifferent arrangements of reservoir parameters can furnish many simulation models tuned to the available past data. #ven though each matched reservoir model is capable of reproducing the observed data, these various geological models can generate different production forecasts. The greatest challenge is obtaining multiple efficient history-matched models for realistic uncertainty estimation. $onventional history matching is a trial-and-error process. The mismatch beteen observed and simulated values is minimi%ed by adjusting reservoir parameters over successive simulation runs. #ven for experienced reservoir engineers, the process is very time consuming and, in general, a single adjusted scenario can be geologically inconsistent. &n the other hand, modern history- matching techni!ues apply numerical optimi%ation and generate multiple geologically consistent adjusted scenarios. The reservoir engineer needs not only to match the past but also has to forecast production by considering various geological models and reservoir development scenarios. 'enerally spea(ing, it consists in finding a set of parameters for the fluid flo e!uations that allos the simulator to reproduce the observed data. More precisely, these parameters could be) * The porosity+ * The absolute permeability, * The relative permeability * The productivity indices of the ells, etc. ,t each ell, the observed data are mainly the pressure, the fluid rate of the different phases and the composition of the fluids. ,ny combination of these variables such as ater-cut, gas-oil and ater-oil ratios can also be used. -n the inversion, it is ell-(non that the anser may not be uni!ue. Hence, it is important to incorporate geological (noledge in the history match procedure to reduce the space of possible solutions. Ta(ing into account geological data, the production forecasts should be more predictive. -n the conventional HM procedure, the main steps are) * .uilding of an -nitial model+ usually, the model is deterministic. The parameters to invert are selected by the reservoir engineer by considering their degree of (noledge and the fluid flo process to simulate. * Matching &bserved "ata+ the observed data are ell pressures, flo rates, etc. To obtain the match, the conventional procedure is done on a try and error procedure, i.e. given a set of parameter, the simulator is run and the results are com-pared ith the observations. /hen this stage is done manually, the reservoir engineer modifies the values of the parameters ith respect of reservoir (noledge and of his understanding of the behavior of the reservoir. -t is sometimes possible to speed up this step by using an automatic procedure hich allos us to modify iteratively the values of the parameters to obtain a better agreement beteen the observations and the computations. * 0imulating 1roduction 2orecast ith Matched Model /hen there is a suitable match of the available data (history match), the same simulator is used to forecast the behavior of the reservoir. 0ometimes, sensitivity studies are done around the parameters obtained after the match, but this does not directly give a !uantification of the uncertainty on the forecasts. 0cenario Matching (2uture 1rediction) The previously discussed method concerns the fit of actual data ith or ithout prior information on the geology. The !uestion of the production forecasting uncertainties obtained ith the different models is never directly addressed. The main steps are) * .uilding of an initial, using the same procedure as in the .- (including all the a priori information) * ,dding data in the future to the observed data to create scenarios+ the choice of these additional data corresponds to tests of hypotheses on production forecasts) for example it may translate a possible ater brea(through at a given future time or hypotheses on oil recovery, etc. * Matching of observed and added data+ for each scenario, the conventional inversion procedure is applied. , ne model (i.e. a ne set of parameters x) is obtained, corresponding to the previous hypotheses. The probability of each scenario is directly chec(ed, using the .- formalism, ith the a posteriori pdf value. * 0imulating production forecasts ith matched scenarios+ for each possible scenario, a direct numerical simulation gives a ne forecast. 0everal cases may occur) * $onsidering the choice of the parameters and the constraints on the models, it is possible to find a set of parameter values to reproduce the measured and the added data. This provides the reservoir engineer ith very useful information. * &n the other hand, the matching of the scenario is only possible by greatly distorting the initial geological model. -n this case, the reservoir engineer is more comfortable ith his prediction. -nversion in 3eservoir #ngineering is often done on the basis of trials and errors runs by the reservoir engineer. This process starts ith an initial guess of the model. 4ihad ,l-4oumaa 5678