张贤松,谢晓庆,康晓东,等.非均质油藏聚合物驱注入参数优化方法与应用[J].深圳大学学报理工版,2018,35(No.4(331-440)):362-367.[doi:10.3724/SP.J.1249.2018.04362]
ZHANG Xiansong,XIE Xiaoqing,KANG Xiaodong,et al.Improvement and application of injection parameters optimization method for polymer flooding in heterogeneous reservoirs[J].Journal of Shenzhen University Science and Engineering,2018,35(No.4(331-440)):362-367.[doi:10.3724/SP.J.1249.2018.04362]
点击复制
非均质油藏聚合物驱注入参数优化方法与应用
张贤松1,2,谢晓庆1,2,康晓东1,2,陈冠中1,2,林春阳1,2

1)海洋石油高效开发国家重点实验室,北京 100028; 2)中海油研究总院有限责任公司技术研发中心,北京 100028

油田开发; 非均质油藏; 聚合物驱; 指标函数; 优化方法; 算法应用

Improvement and application of injection parameters optimization method for polymer flooding in heterogeneous reservoirs
ZHANG Xiansong1,2, XIE Xiaoqing1,2, KANG Xiaodong1,2, CHEN Guanzhong1,2, and LIN Chunyang1,2

1)State Key Laboratory of Offshore Oil Exploitation, Beijing 100028, P.R.China2)Research Centre of Research Institute Co. Ltd., CNOOC, Beijing 100028, P.R.China

oilfield development; heterogeneous reservoir; polymer flooding; index function; optimization method; algorithm application

DOI: 10.3724/SP.J.1249.2018.04362

备注

聚合物驱注入参数优化是提高非均质油藏聚合驱技术效果和经济效益的关键技术之一.聚合物驱注入参数优化属于复杂的大规模最优化问题,目前仍缺乏与之相适应的高效求解算法,严重限制了注入参数优化方法的研究与应用.本研究通过理论推导得到由有限差分随机逼近方法(finite difference of stochastic approximation, FDSA)引导的随机扰动逼近改进新算法(simultaneous perturbation stochastic approximation algorithm, SPSA-FDG),可提高计算收敛速度,在迭代计算过程中,各控制变量近似梯度大小比例与FDSA算法类似,梯度估计精度高于SPSA算法,并保留了SPSA算法每次迭代计算次数少的优点.实例应用表明,SPSA-FDG算法迭代收敛时需要248次油藏数值模拟计算,明显少于FDSA算法和SPSA算法.此外,在聚合物驱注入总量不变条件下,通过应用改进算法和油藏数值模拟手段,优化聚合物驱每一段塞下的最优注入浓度和段塞尺寸,得到的聚合物驱优化注入方案净现值比均匀注聚方案提高11.64%,可使油田开发技术经济效果最大化.

The optimization of injection parameters is the key to the improvement of both the technical and economic performances of polymer flooding in heterogeneous reservoirs. This injection parameter optimization is a kind of large-scale complicated optimization problem, yet a lack of efficient solutions. In this paper, a new method is proposed using the simultaneous perturbation stochastic approximation algorithm(SPSA-FDG)led by the finite difference of stochastic approximation(FDSA). Thus, the approximated gradient ratio of each control variable is similar to that of FDSA algorithm in the iteration process. The studies show that the precision of gradient estimation using SPSA-FDG algorithm is higher than that using SPSA algorithm, and the convergence of SPSA-FDG algorithm is better than that of FDSA and SPSA algorithm. In addition, the SPSA-FDG algorithm could also keep the advantages of SPSA algorithm which undergoes only twice reservoir numerical simulations at each iteration. Under the condition of same polymer injection amount, the net present value of optimal polymer flooding injection scheme by the SPSA-FDG algorithm is 11.64% higher than that of uniformity design, which helps achieve the better economic performance of polymer flooding.

·