Optimizing the performance of smart wells in complex reservoirs using continuously updated geological models

Optimizing the operation of smart wells (i.e., wells with downhole chokes and sensors) remains a challenge, particularly in light of geological uncertainty. In this paper a combined valve optimization and history-matching procedure is presented and applied. The overall approach allows for the continuous updating of the geological model using data inferred from downhole sensors. The method uses numerically computed gradients, generated using a commercial reservoir simulator for function evaluations, for the valve optimization in conjunction with a probability perturbation approach for the history matching. The methodology is applied to three examples involving both variogram and multiple point geostatistical models. For these cases, using the proposed methodology, oil production is shown to be nearly equal to that achieved using optimized valves with known geology, indicating the potential benefits of the overall approach. In one of the cases, optimization over multiple history-matched models provides significant improvement over optimizing with a single model.