Optimizing In-fill Well Placement in a Mature Waterflood Field Using a Physics Embedded Machine Learning Workflow

In-fill drilling is a common IOR practice to increase production in mature fields; operators target oil that was not effectively swept by the water injection by reducing well spacing and increasing the interaction between injectors and producers, particularly in heterogenous reservoirs. In-fill drilling can improve overall recovery, however if the wells are not properly located, the practice may end up in increased capex and lifting cost and not render the expected results.

Using traditional reservoir simulation to optimize in-fill well placement is not necessarily an efficient way to do it. Usually, reservoir engineers will simulate placing some in-fill wells in places where they believe they know the results would be acceptable and compare the different results before recommending drilling locations. Ideally one would like to “virtually” drill every possible location as defined by the drainage area of location density- and evaluate the impact of each well separately.

In large fields, this means running thousands of possible locations, a task not necessarily achievable considering the speed at which simulation models run.

This paper presents a methodology to optimize in-fill drilling location selection using Data Physics, a combination of traditional reservoir modelling physics, machine learning and advanced optimization techniques.