Case Study: Agave achieves weekend-scale probabilistic forecasting with Stone Ridge Technology’s ECHELON GPU simulator

Agave Highlighting Echelon Software'S Robust Compositional Solver

Agave, a light-oil deep-water field with early production history, needed a probabilistic forecast and risk analysis for a proposed development strategy based on water and gas reinjection. The challenge was to run an ensemble of 150 history-matched compositional models with 2.4 million active cells and 11 hydrocarbon components per region fast enough to complete over a weekend on a limited share of Eni’s HPC5 supercomputer. Stone Ridge Technology provided ECHELON 2.0 to address this need.

Stone Ridge Technology used ECHELON 2.0, its GPU-based compositional simulator, to run the forecasts efficiently in fully implicit mode, with a best compromise of one node per model using four GPUs. The results were robust across GPU configurations, and the workflow completed the full Monte Carlo risk analysis over the weekend, with most runs finishing in under four hours and the slowest outliers taking up to ten hours. This enabled Agave to deliver cumulative oil production distributions and technical reserves estimates from the 150-case ensemble.