Development of Effective Carbonate Steamflood Strategy Using Full-field Simulation Models and Machine Learning Algorithms

Summary Steamflooding is a widely used thermal method for recovering heavy oil from sandstone reservoirs. In carbonates, the implementation of steamflooding usually demonstrates higher steam-oil ratio and lower oil recovery. The key performance problem is a poor sweep efficiency of steam injection. It is fully confirmed by actual results of steamflooding in the Permian – Carboniferous carbonate reservoir of the Usinsk field located in Northwest European Russia. The reservoir has the largest heavy oil remaining reserves in carbonates of Russia and Europe. Since the viscosity of its oil is more than 700 mPa*s, in some areas of the reservoir, there is a steam injection at ~300°C and ~10 MPa, which are being used for almost 40 years mostly via vertical wells. However, the current oil recovery numbers of the areas are estimated only between 12 and 15 %. It is assumed that these oil recovery efficiencies could be improved with optimized reservoir management with advanced numerical modeling to evaluate the additional oil production and steam-oil ratio and figure out the best further steamflooding strategy. For many years, an exclusively deterministic approach was used to simulate the reservoir, which significantly limited the possibilities for modifying the steam injection process. That is why, the search for alternative approaches of reservoir modeling, which ensure prompt obtaining realistic forecasting of its development, was relevant. In this work, a novel forecasting technology termed an adaptive approach that combines the full-field geological and hydrodynamic models with the unique machine-learning algorithm based on fuzzy-logic functions was implemented. The obtained results of the adaptive approach application demonstrated the improvement in understanding of the reservoir thermal performance and in making the practical recommendations of cost saving and oil production increase.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.