聚合物驱注采参数无梯度优化新算法

A new algorithm for non-gradient optimization of injection-production parameters for polymer flooding

无梯度优化算法对最优化算法理论研究具有重要意义,已经被应用于许多实际问题的解决,也可以用来求解聚合物驱注采参数最优控制模型问题。针对有限差分随机逼近(FDSA)算法和同时扰动随机逼近(SPSA)算法的优缺点,提出了一种有限差分梯度引导的同时扰动随机逼近(SPSA-FDG)算法,建立了油藏数值模拟模型,开展了聚合物驱注采参数最优控制研究,并将3种算法的结果进行了对比和分析。结果表明:SPSA-FDG算法收敛性要好于FDSA算法和SPSA算法,且易于和任何油藏数值模拟器相结合,能够用来求解聚合物驱最优控制问题;与均匀注聚方案对比发现,采用由SPSA-FDG算法计算得到的优化方案,聚合物驱见效时间提前,含水下降漏斗加深,累积采出程度提高,油藏总体开发效果更好。SPSA-FDG算法可应用于其他化学驱油田以改善化学驱效果。

Non-gradient optimization algorithm for optimization algorithms theory research is meaningful and has been applied in many practical problems,and it can also be used to solve the optimal control model of injectionproduction parameters for polymer flooding. Aiming at the advantages and disadvantages of finite difference stochastic approximation(FDSA)and simultaneous perturbation stochastic approximation(SPSA),a simultaneous perturbation stochastic approximation algorithm guided by finite difference gradient(SPSA-FDG)was proposed.Based on CMG reservoir numerical simulation software,reservoir numerical simulation model was established,by using the SPSA-FDG,optimal control study of injection-production parameters for polymer flooding was carried out,and the optimized results with the three algorithms were analyzed. The results show that the convergence of SPSA-FDG algorithm is better than that of FDSA and SPSA algorithm,and it is easy to combine with any reservoir numerical simulator to solve the optimal control problem of polymer flooding. Compared with uniform polymer flooding scheme,it is found that by using the optimized scheme calculated by SPSA-FDG algo-rithm,polymer flooding efficiency time is ahead of schedule,the water cut drop funnel is deepened,the cumulative recovery degree is improved,and the overall reservoir development effect is better. SPSA-FDG algorithm can be applied in other chemical flooding fields to improve the effect of chemical flooding.