Predicting Fluid Production Rates in Tight Carbonates: How Ensemble Learning Links Static and Dynamic Data

Summary In naturally fractured tight carbonate reservoirs, estimating the potential productivity of a given interval is a longstanding problem, mainly due to the inherent multi-scale heterogeneities to be modeled. Conventional approaches, usually based on the prediction of permeability logs, are not an option due to the difficulty in obtaining reliable petrophysical characterizations. A predictive analytics methodology for estimating fluid production in tight carbonate reservoirs is proposed, discussed, and validated with real case applications. The approach uses static data from micro-resistivity image logs and from other geological/drilling information as inputs, together with dynamic well performance as the target.