基于代理模型与遗传算法的海上油田聚合物驱注入参数优化控制方法

Optimizing injection parameters of polymer flooding in offshore oilfields by combining surrogate model and genetic algorithm

针对传统优化方法效率低、精度差的问题,将响应面方法引入到化学驱注入参数优化设计中,提出了基于二次响应面代理模型与遗传算法的海上油田聚合物驱注入参数优化控制方法,即运用Box-Behnken试验设计方法,以油藏数值模拟为计算手段,采用二次响应面方法建立了拟合净现值增幅与注入参数之间非线性关系的代理模型,并以此作为适应度函数,由遗传算法在变量空间范围内进行全局寻优,获得注入参数最优组合。渤海SZ36-1油田A7井组聚合物驱注入参数优化结果表明,基于代理模型和遗传算法的优化策略可以有效减少数值模拟的次数和计算量,同时能够获得较高的求解精度。

The conventional methods to optimize injection parameters of polymer flooding are quite lower in efficiency and accuracy, and a new optimization method by combining a surrogate model based on quadratic response surface with genetic algorithm was presented, after the response surface tool was introduced into the design of injection parameter optimization for offshore polymer flooding. In other words, the response surface tool was applied to build a surrogate model to fit nonlinear relationships between the increments of net present value and the injection parameters, by taking Box- Behnken test as a design tool and the numerical simulation of oil reservoirs as a calculation tool. Then this surrogate model was used as a adaptability function to make a complete optimization searching within the space of variables, leading to the optimized combination of injection parameters. This method was applied to well group A7 in SZ36- 1 oilfield in Bohai bay, and the results showed that the optimization by combining a surrogate model with genetic algorithm could effectively reduce nu merical simulation times and calculations and obtain a higher accuracy of solution.