随机系数下线性二次控制问题的最优性条件及其应用

本文旨在研究随机系数下随机微分方程的线性二次最优控制问题.本文从闭环最优控制/策略存在的必要性条件的角度开展研究.若闭环最优控制/策略存在,得到其显示反馈表示、带伪逆运算的倒向随机Riccati方程的适定性及不同系数间满足的一些本质性条件.此处结论本质地推广和改进了文[Ait Rami M,Moore J,Zhou X.Indefinite stochastic linear quadratic control and generalized differential Riccati equation[J].SIAM J Control Optim,2001,40:1296-1311;Sun J,Yong J.Linear quadratic stochastic differential games:open-loop and closed-loop saddle points[J].SIAM J Control Optim,2014,52:4082-4121;Lü Q,Wang T,Zhang X.Characterization of optimal feedback for stochastic linear quadratic control problems,Probab Uncertain Quant Risk,2017,2:11,DOI 10.1186/s41546-017-0022-7]的相应结论.此外,本文得到了一个关于倒向随机Riccati方程和二阶伴随方程两类方程适应解之间的微妙关系.注意到,这一结论在现有文献中首次出现.最后,本文讨论了在均值方差对冲问题中的应用.

Theoretical research on reservoir closed-loop production management

Closed-loop production management combines the process of history matching and production optimization together to peri-odically updates the reservoir model and determine the optimal control strategy for production development to realize the goal of decreasing the knowledge of model uncertainty as well as maximize the economic benefits for the expected reservoir life. The adjoint-gradient-based methods seem to be the most efficient algorithms for closed-loop management. Due to complicated calculation and limited availability of adjoint-gradient in commercial reservoir simulators, the application of this method is still prohibited for real fields. In this paper, a simultaneous perturbation stochastic approximation (SPSA) algorithm is proposed for reservoir closed-loop production management with the combination of a parameterization way for history matching and a co-variance matrix to smooth well controls for production optimization. By using a set of unconditional realizations, the proposed parameterization method can transform the minimization of the objective function in history matching from a higher dimension to a lower dimension, which is quite useful for large scale history matching problem. Then the SPSA algorithm minimizes the objective function iteratively to get an optimal estimate reservoir model. Based on a prior covariance matrix for production op-timization, the SPSA algorithm generates a smooth stochastic search direction which is always uphill and has a certain time correlation for well controls. The example application shows that the SPSA algorithm for closed-loop production management can decrease the geological uncertainty and provide a reasonable estimate reservoir model without the calculation of the ad-joint-gradient. Meanwhile, the well controls optimized by the alternative SPSA algorithm are fairly smooth and significantly improve the effect of waterflooding with a higher NPV and a better sweep efficiency than the reactive control strategy.