Using lower-fidelity models to accelerate production strategy optimization of a pre-salt carbonate reservoir

Abstract Reservoir simulators are essential for forecasting and production optimization, but their use in complex fractured reservoirs is often limited by high computational cost. This work proposes a streamlined method for building lower-fidelity, fit-for-purpose models (LFMs/FPMs) that preserve key reservoir behaviors while significantly reducing runtime. In this work, we found that, while a typical lower-fidelity model is faster and has lower reliability, a chosen FPM keeps close reliability and significantly higher performance than the original HFM (higher fidelity model) or full-physics model within quantitative acceptance of model properties. Unlike traditional upscaling, the approach simplifies geological and petrophysical inputs, such as permeability and porosity, without altering the grid structure, thus minimizing geomodelling effort. This study demonstrates that input and property-based simplification and quality control can provide a practical and accurate alternative to high-fidelity simulation, offering the offshore E&P industry a powerful tool for faster, more efficient decision-making. Applied to a carbonate reservoir in southeast Brazil, the method delivered substantial speedup in production-strategy optimization while maintaining high agreement with the full-physics model, 94% speedup with an equivalent response quality, R 2 above at least 99% on all criteria, and correlation coefficient above 79% on the criteria that matter most. The LFMs reproduced decision-relevant responses with strong consistency across multiple metrics, supporting their use in robust optimization workflows. The final development strategies derived from the FPMs matched or slightly outperformed FEI from the original model well positioning optimization.