Optimization of Injection Well Placement for Waterflooding in Heterogeneous Reservoirs using Artificial Neural Networks Coupled with Reservoir Simulation

Secondary recovery methods such as waterflooding and gasflooding are often applied to depleted reservoirs for enhancing oil and gas production. Reservoir simulations are performed to predict the hydrocarbon production by secondary recovery methods in heterogeneous fields. Given that a large number of discretized elements are required in simulations, it is usually not technically-and-economically feasible to run full-physics simulation for every possible case. In this regard, machine learning technology is introduced to predict the hydrocarbon production efficiently. In this paper, we firstly review the previous works on the heterogeneous reservoir simulations and the applications of Artificial Neural Network (ANN) models to predict the reservoir responses. Secondly, we present an improved data-driven models using ANN for the prediction of hydrocarbon production by waterflooding in heterogeneous reservoirs. It is proposed to improve the prediction performance while reducing the monitoring cost. Injection well placement can be optimized by using the proposed ANN models.