The simplest and oldest method to represent hydraulic fractures in reservoir simulation models is to use a negative skin factor for stimulated wells, the implicit representation. However, a negative skin factor approach does not capture several flow aspects that impact production significantly, especially in heterogeneous, lower permeability reservoirs. A better, more realistic, production forecast is obtained by modeling the effects of induced fractures and their associated flow patterns explicitely in the dynamic reservoir model using local grid refinements (LGR). This approach has been used and demonstrated in the industry for several decades and has proven to be a reliable and flexible approach. In single well models, with current computing power, this approach works well. In full field models, however, dynamic reservoir models with LGR's are computationally very demanding, and not applicable in many cases.To overcome these computing problems in larger, more complicated models with many wells, a so-called proxy model was developed that also represents a fracture explicitly. In the proxy approach, the fracture is modelled using a set of well completions placed in the host gridblocks associated with the (planned) hydraulic fracture geometry. The explicit proxy model approach is significantly (10x) more efficient in run-times, allowing for much faster scenario developments, while maintaining a similar quality production forecast.To calibrate the quality of the production forecast from the proxy approach, the proxy model settings were tuned against the LGR approach. An important aspect of this is in setting the transmissibility from the reservoir to the completions to match the productivity calculated with an explicity LGR model. The (local) PI of a proxy completion is controlled by a so-called proxy connection (transmissibility) factor. Proxy connection factors, in the Proxy approach, are a function of; dimensionless fracture conductivity, dimensionless fracture height, dimensionless fracture length, permeability heterogeneity effects and selected grid discretization options.To validate the Proxy model for a real field development scenario, the Proxy approach was assessed against the LGR approach for three wells in a very heterogeneous gas condensate reservoir. The LGR models were developed by history matching several years of actual production history, and then used to make forecast simulations. For all three wells, a close match was obtained between productivity forecasts of the Proxy and the LGR approach.Next to the improvement in efficiency, a further advantage of the Proxy model is that the fractures are defined in the schedule section of the run. In full field production forecasting, this allows for the representation and activation of a fractured well later in the life of the field, without disturbing the production simulation before the specific wells and fractures are present in the field, which is an issue with the LGR approach.
