Injector-producer interaction, that is, how well a producer is connected to each of its surrounding injectors, is used in designing efficient injection schemes and infill drilling programs. Determining well interaction in a field with multiple injectors and producers is a challenge because of the heterogeneity of the reservoir. In addition, the presence of subseismic faults, both conducting and nonconducting, further curtails our ability to estimate well interaction indices.
Currently, the injector-producer interaction is estimated by visually cross correlating injection and production rates of different well pairs in a pattern. Regional discontinuities in reservoir properties including faults and pinchouts are inferred from the cross correlation. Such statistics-driven methods are stable as shown previously in many such fields of applications. There are, however, two problems with this approach: firstly, the results obtained using the cross correlation are non-unique and secondly, the process is extremely time consuming.
This paper presents a new integrated approach to determine the injector-producer interaction. First, a multi-variate data set consisting of production, injection, petrophysical, sand/shale, and well location information is generated. An artificial neural network (NN) is then trained to estimate the well interaction between different well pairs. The estimated well interactions are used to determine the presence of heterogeneities, such as faults, pinchouts, regional permeability trends etc. Results show that the new integrated app roach can quantify injector-producer connectivity more accurately, consistently, and inexpensively than the conventional methods. The new method will, thus, facilitate better reservoir management of fields where the knowledge of injector- producer interaction can affect recovery efficiency, sweep, reservoir performance, and infill well placement.