Application of brown-field experimental design techniques to a super giant carbonate reservoir

Brown-field experimental design techniques were applied to the Tengiz super-giant carbonate reservoir. This brown-field study combined experimental design, reservoir simulation, and available historical pressure and production data to develop proxies to predict the history match quality and oil recovery of equally probable reservoir realizations. The authors used experimental design techniques similar to those documented in Landa et. al.¹ and King et al.²; however, the effort documented in this paper incorporated a much larger brown-field component and highly non-linear response surfaces required many more runs (control points) to generate reliable proxies. The initial simulations runs included a folded Plackett-Burman design with a centerpoint run and sensitivity analysis runs which varied the input parameters that were known to affect the history match. The results from these runs were analyzed using regression analysis and ANOVA techniques to determine the most significant factors and to understand the complexity of the response surfaces. With this knowledge, D-optimal runs were designed and run to better understand the effect of the interaction of the significant factors on the response surfaces. Additional sensitivity analysis runs were also made to further define the effect of the two most important factors on the response surfaces: OOIP and reservoir transmissibility. With a total of 107 unique runs, the authors were able to develop five proxies to assess history match quality and one proxy to predict oil recovery. With the developed proxies, the authors used Monte Carlo techniques to generate equally probable reservoir/development scenarios. By using the history match quality proxies the authors were able to filter out combinations of subsurface parameters that caused unacceptable deterioration in the history match. The use of the five unique filters allowed the authors to significantly reduce uncertainty relative to the unfiltered simulation results. The results from this work were supported by material balance studies.