Hierarchical optimization in short-term and long-term of reservoir management through surrogate models

Jefferson Wellano Oliveira Pinto¹; Silvana Maria Bastos¹; Ramiro Brito Willmersdorf¹

¹ Universidade Federal de Pernambuco

doi:10.20906/CPS/CILAMCE2017-0273

Resumo

Optimization techniques have been extensively used to achieve feasible and economical designs in most varied fields of engineering. Nowadays the approaches have become increasingly realistic, having been commonly used to solving non-trivial problems of practical engineering, including the optimum management of reservoirs through computational simulation. One point only recently tackled concerns the short-term objective, which in most production optimization studies is neglected. As a result, the production and injection flow obtained from optimization often results in a considerable reduction of short-term production performance. And it is generally these objectives that define the course of the operational strategy, especially due to reservoirs geological and economic uncertainties. From the operational perspective, the decision on how to operate the wells is quite critical, since the long and short term optimization strategies may greatly differ. Each option has pros and cons, the strategy of producing more oil in the short term can leave an appreciable amount of oil in the reservoir, however this option is not so affected by the present uncertainties, such as in the oil price that occurs in a long term production. To solve the problem, a hierarchical optimization structure with multiple objectives is adopted. In this work two approaches are used. The first one considers only the NPV as objective function and in its calculation different annual discount rates are used as a parameter to emphasize a short or long term optimization. In the second approach, NPV is optimized with a fixed discount rate and with the solution obtained, a least squares method is applied on oil production in a short pre-established time interval. Two reservoir problems are analyzed, both are black-oil models and they are simulated by CMG's commercial IMEX simulator. To reduce the computational cost of simulation, surrogate models based on data fitting using radial basis functions (RBF) are constructed. In this case the surrogate model

Palavras-chave: Optimal management; Hierarchical optimization; Short-term, long-term optimization; Sequential Approximation Optimization