Integrated modeling and parallelized optimization of enhanced oil recovery processes in hydrocarbon reservoirs under uncertainty

由于新发现的油田在过去几十年里不断减少，目前世界石油产量的相当一部分来自采用提高采收率（EOR）技术的成熟油田。在成功实施EOR工艺之前，通过将静态地质模型与动态生产数据进行校准，准确确定成熟油田中残余油的数量和位置至关重要，然后可以优化生产注入策略。由于模拟器的输入与生产数据、所采用的模拟方法、所收集的数据和模型规模之间存在复杂的非线性关系，通常会降低储层地质模型的准确性。地质和经济不确定性的存在，不仅增加了准确、动态更新地质模型的难度，而且将优化难度提升到风险评估问题的高度。因此，根据静态和动态数据生成多种油藏模型，并在不确定条件下对相应的油藏动态进行优化，具有基础性和现实意义。

A considerable portion of current world oil production comes from mature oilfields by applying enhanced oil recovery (EOR) techniques since new discoveries have been declining steadily over the last few decades. Prior to successfully implementing an EOR process, it is essential that the amount and location of the residual oil in a mature oilfield be accurately determined by calibrating a static geologic model to dynamic production data and then the production-injection strategy can be optimized. Due mainly to the complex and nonlinear relationship between inputs of a simulator and production data, simulation methods used, data collected, and model scale, accuracy of a reservoir geological model is normally compromised. The presence of geological and economic uncertainty not only makes it more difficult to accurately and dynamically update the geological model, but also elevates the difficulty of optimization to the level of a risk assessment problem. Therefore, it is of fundamental and practical importance to generate multiple reservoir models conditional to static and dynamic data and subsequently optimize the corresponding reservoir performance under uncertainty.