Rank Reservoir Connectivity Using Dynamic Data

Tang, Hong¹, wei Xu² (1) Louisiana State University, Baton Rouge, LA (2) Louisiana State University, Baton Rouge, LA

Even though production data have direct responses of reservoir heterogeneity and connectivity, they are rarely incorporated into reservoir modeling workflow. In this paper, an experiment-designed, probability perturbation method (EDPPM) is proposed to quantify the uncertainties of reservoir connectivity. This method proposes to be more accurate and efficient by integrating both static and dynamic data. It is divided into 3 steps and has been developed with a synthetic shore-face-fluvial reservoir. 1. Multiple typical reservoir realizations with four geological factors (NTG, fault sealing, OWC and tar mat) are generated by a Plackett-Burman design. Geostatistic modeling realizations calibrated with well data and geological prior probability represents geological uncertainties. 2. The connectivity factor is defined as a function of recovery factors from flow simulations. The designed reservoir geostatistic realizations are treated as initial models. Multiple point statistics method is used to locally adjust the geological facies distribution. A probability perturbation method computes and minimizes the difference between simulation results and production data by adjusting modeling parameters. 3. A linear response surface is modeled. Monte Carlo simulation generates posterior probability distributions of connectivity factors and OOIP. Both distributions provide P10-50-90 models for business decision. Analysis of variance shows EDPPM identifies OWC and interaction term have significant impact on connectivity, which is more accurate than pre-perturbed design. Furthermore, when the NTG is high, facies connectivity will have less impact on oil recovery; and EDPPM is easier to converge, vice versa. EDPPM honoring dynamic data is more accurate for risk mitigation with acceptable computational cost.