Abstract
Rigorous and continual advancement in the computational domain have drastically reduced the time consumed for simulations. Despite this, the heterogeneity and requirement to use compositional models, among constraints such as field size, can still thwart the optimization of the field development plan (FDP) in terms of time spent. To address this, we proposed and compared different techniques to underscore how to optimize FDP efficiently under uncertainty using an example of a giant benchmark field. We compared four workflows to improve the efficiency of the optimization process. Method 1 uses a full-field model (FFM) approach with an intelligent selection process to eliminate the poorest FDPs. As it is usual to divide a field into sectors for risk-aversion and strategic purposes, Method 2 uses an isolated-sector approach to reduce average simulation time. Method 3 employs the FFM approach to perform the optimization using only the monetary value of the partial life of the field. Method 4 uses a cluster-based search space reduction technique for a predictive analysis from the technical results obtained with partial simulations, which are similar to Method 3. To ensure good decisions, the optimized FDP was always implemented in the FFM with complete field-life at the end of all methods. All proposed workflows are promising in terms of efficiency, acknowledge the entire envelope of uncertainty, and consider multiple scenarios to improve the chances of success of the optimized FDP in the real field. Aside from obtaining good results when compared to traditional methods, we also saved 80-93% of the computational time with these methods. Thus, one can reduce exorbitant costs and delays in performing the FDP optimization. As anticipated, we observed a generic trade-off between decreasing computational time and increasing field objective function. Despite this, using a new algorithm with predictive analytics in Method 4 produced the best improvement within the shortest timespan, which demonstrates that one can use shorter-term data to understand a field's non-linear response. Numerous authors have already presented various algorithms to optimize an FDP. Despite the existing computational capabilities, these algorithms are still inept at developing a field with time-consuming simulation models. Unlike previous studies, this work presents practical solutions to assist field development, considering probabilistic scenarios to capture associated uncertainty. We also discuss the advantages and disadvantages of the methods to establish their application in different situations.
NSTL