Production Acceleration for a Mature Waterflood in Kuwait through Predictive Model Combining ML with Reservoir Physics

The paper describes the application of Data Physics (physics embedded machine learning) to model and optimize a large mature field in the province of KOC West Kuwait Asset. The field has been in production since 1959 and water-flooding started in 2003. The field has currently close to 100 active producers and 15 injectors. After creating the Data Physics model, an optimization workflow using evolutionary algorithms was run to generate injection optimum prescription, resulting in a potential production increase of 5% maintaining the current injection levels and hence with no significant increase of cost.