Efficient well placement optimization coupling hybrid objective function with particle swarm optimization algorithm

Abstract

Well placement optimization is a critical part of the oil field development planning which aims to find the optimal locations of wells to maximize a traditional objective function (TOF), e.g., cumulative oil production (COP) or the net present value (NPV). However, the optimization process can be quite time-consuming since it requires iterative evaluations of the objective function and each evaluation requires one simulation run of the fluid flow in the discretized time domain and space domain. This paper examined the consistency between productivity potential value (PPV) and cumulative oil production (COP), and proposed to use PPV as the objective function, whose evaluation does not require simulation run, to improve computational efficiency. However, since PPV is a static measure of a reservoir, we use PPV only in early iterations of the well placement optimization followed by TOF in later iterations in order to capture reservoir dynamics. The use of PPV and TOF in a sequential manner is referred to as a hybrid objective function (HOF). In this work, a naturally parallelizable optimization algorithm, particle swarm optimization (PSO), where simulation runs can be conducted in batches is used as the optimizer. The effectiveness of the proposed procedure is validated based on three numerical examples including a 2D model, the PUNQ-S3 model and the Egg model. Results demonstrate the well placement optimization strategy using HOF finds comparable COP within much less simulation runs compared to the optimization using TOF. In summary, well placement optimization with the objective function defined as PPV in the first 25% iterations and TOF in the following 75% iterations is the best combination.