Inflow control device (ICD) restricts flows of certain sections to achieve a uniform push of water oil contact along a horizontal well. The efficiency of ICD depends on its configuration (location, number of nozzles, and nozzle sizes). Traditionally, optimal ICD configurations have been selected using a trial and error approach simulating different configurations. This method is time-consuming and the outcome can be limited by pre-set candidates, thus may result in non-optimal configurations. In this paper, a method involving several optimization objectives is proposed to obtain the optimal ICD configuration, in order to reach an optimal total production before water breakthrough. A flow coupling model (permeation flow – nozzle flow – inner tubing flow) of horizontal wells completed by ICDs is established to calculate the inflow profile and the pressure distribution along the horizontal wellbore. The production rate and the water breakthrough time are the optimization objectives, and the sand control condition (production pressure drawdown < critical pressure drawdown of sanding onset) is taken as one of the constraints of optimization. Then the NSGA-II algorithm is employed to search for the Pareto frontier based on the model of coupled flow. Selected by the TOPSIS method from the Pareto frontier, the optimal ICD configuration ensuring optimal total production and no sanding is acquired. The original data of well A1 in SL oilfield are used for case study. The optimal configuration is compared with single-objective optimization results and open hole completion in terms of production rate and water breakthrough time. By predicting the total productions of different configurations before water breakthrough, we think the optimization method is capable of determining the optimal configuration. This research can provide some theoretical support for the determination of ICD configuration in horizontal wells.

流入控制装置（ICD）限制某些部分的流量，以实现水油接触沿水平井的均匀推动。ICD的效率取决于其配置（位置，喷嘴数量和喷嘴尺寸）。传统上，使用模拟不同配置的反复试验方法来选择最佳的ICD配置。该方法很耗时，结果可能会受到预设候选对象的限制，因此可能导致配置不理想。本文提出了一种涉及多个优化目标的方法，以获得最佳的ICD配置，以便在水突破之前达到最佳的总产量。建立了由ICD完成的水平井的流耦合模型（渗透流–喷嘴流–内管流），以计算沿水平井眼的流入剖面和压力分布。生产率和水突破时间是优化的目标，防砂条件（生产压力下降<喷砂开始的临界压力下降）被视为优化的约束条件之一。然后，基于耦合流模型，使用NSGA-II算法搜索帕累托边界。通过TOPSIS方法从帕累托边界中选择，最佳的ICD配置可确保最佳的总产量，并且不会打磨。SL油田A1井的原始数据用于案例研究。将最佳配置与单目标优化结果和裸眼完井的生产率和水突破时间进行比较。通过预测水突破之前不同构型的总产量，我们认为优化方法能够确定最佳构型。该研究可以为水平井中ICD构型的确定提供理论依据。