CO2 EOR Performance Evaluation in an Unconventional Reservoir through Mechanistic Constrained Proxy Modeling

Abstract

There are multiple parameters including reservoir characteristics, hydraulic fracture design, injection solvent selection, and the enhanced oil recovery (EOR) operational design, etc., that are directly and indirectly involved in the unconventional reservoir development. And due to the lack of field data availability, the key performance indicators for such complex reservoirs are not well defined, therefore, reservoir engineers end up with a huge volume of numerical simulation scenarios to find an optimum field development plan. This approach makes the field development planning process complicated and time-consuming.In this study, a smart unconventional EOR performance evaluation tool is presented utilizing intelligent modeling strategies through proxy modeling approach with mechanistic constraints. The mechanistic constraints incorporate physics-based numerical simulation dataset. Key reservoir and hydraulic fracture topographies including average reservoir pressure and the matrix & hydraulic fracture permeability contrast are used as some of the key input parameters for the Deep Neural Network (DNN) modeling. Total four DNN models are generated for primary and EOR cases presenting the recovery performance prediction at two different timelines. The trained and validated DNN models are finally found well qualified to be applied for the development of the unconventional reservoirs with identical reservoir and hydraulic fracture characteristics for the EOR application forecasting.