基于数值模拟的油藏生产动态优化算法

Reservoir Production Performance Optimization Algorithm Based on Numerical Simulation

对于大规模油藏开发优化,常规优化算法存在收敛速度慢、寻优效率低、难以与现场结合等问题。针对这些问题,建立了油水井动态调控优化模型,并结合模拟退火遗传算法和拉丁超立方采样算法寻找模型全局最优解,同时,利用同步扰动随机逼近算法提高了模型局部求解收敛速度,研发编制了油水井动态调控优化软件,并应用于大庆油田H区块。对比油田常规油水井生产制度,油水井动态调控优化模型的最优方案使H区块累计产油量在5年内增加了5.68×10^(4)m^(3),较好地解决了油水井生产动态控制优化问题,也为大规模油田高效开发提供了一种新的方法。

When the conventional optimization algorithms are applied to optimized development of large scale reservoirs,the problems such as slow convergence speed,low optimization efficiency and difficult integration with field applications occur.To solve these problems,a well production performance control model was established.A global optimal solution of the model was found by using the simulated annealing genetic(SAG)algorithm and Latin hypercube sampling(LHS)algorithm.Furthermore,the convergence speed of the local solution of the model was accelerated by using the synchronous perturbation stochastic approximation(SPSA)algorithm,and a well production performance control software was developed and applied to the H block in Daqing oilfield.Compared with conventional well production systems,the best scheme of the optimized well production performance control model increases the cumulative oil production of H block by 5.68×10^(4)m^(3)within 5 years,which ensures the well production performance control and optimization,and provides a new method for efficient development of large-scale oilfields.