https://hal-ifp.archives-ouvertes.fr/hal-02284357Metla, N.N.MetlaIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesDelbos, FrédéricFrédéricDelbosIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesda Veiga, SébastienSébastienda VeigaIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesSinoquet, DelphineDelphineSinoquetIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesConstrained nonlinear optimization for extreme scenarii evaluation in reservoir characterization.HAL CCSD2010[SDU] Sciences of the Universe [physics][SDU.STU] Sciences of the Universe [physics]/Earth SciencesMallaret, Pascale2019-09-11 16:42:292021-01-12 13:58:022019-09-11 17:00:24enConference papersapplication/pdf1The goal of reservoir characterization is the estimation of unknown reservoir parameters (the history matching problem), by integrating available data in order to take decisions for production scheme and to predict the oil production of the field in the future (the forecast problem). The reservoir parameters could be classified in two classes: • those related to the geological modeling (spatial distribution of porosity, permeability, faults), • and those related to the fluid flow modeling (relative permeability curves, productivity index of the wells). Those parameters could not be directly determined by measurements (or only locally using well logs), this is the reason why this parameter estimation problem is formulated as an inverse problem with forward simulators that compute synthetic measurable data from those parameters. Observed data are well data acquired at production/injection wells (bottom-hole pressure, gas-oil ratio, oil rate) at different calendar times during the production of the field. The main contribution of this work is the integration of nonlinear optimization methodology to predict the oil production of a field and to give a confidence interval on this prediction. We believe that applying non linear optimization methods will increase accuracy and then give more reliable production forecast than approaches with simplified models of forward operators (linear approximations or response surfaces). The first and second sections of this paper are respectively dedicated to the history matching problem and to the forecast problem. In the third section, we described the optimization methods used to solve both problems. Then, in the last section the previous methodology is applied to a 3D synthetic reservoir application (the PUNQ test case).