基于主成分分析和代理模型的油藏生产注采优化方法

A reservoir production optimization method based on principal component analysis and surrogate model

针对油藏注采优化变量多,直接使用代理模型优化结果变差的问题,提出基于主成分分析和代理模型的油藏生产注采优化方法,代理模型通过对油藏数值模拟器采样,构建速度快、逼近原数值模拟模型精度的数学模型。通过主成分分析对数据进行降维,而后借助基于代理模型的差分进化算法进行优化,最终得到实际生产优化问题的调控方案,并应用SADE-PCA方法对Egg模型进行实例计算。结果表明:提出的基于主成分分析的代理辅助进化算法,在保证精度的前提下,可有效降低变量的维数,解决油藏大规模变量注采优化的难题;优化的注采方案能够有效增油控水,提高油田的经济效益;插值控制方法(ICM)在实现平滑的调控方案和提高计算效率方面具有良好性能。

When dealing with reservoir production optimization problems with many variables,directly using surrogate model may deteriorate the optimization results.In this study,a surrogate-assisted evolutionary algorithm based on principal component analysis was proposed for production optimization.Surrogate model is a mathematical model that is computationally fast with an accuracy close to the original numerical simulation model.The principal component analysis was used to reduce the dimensionality of the data,and then it was optimized by using a differential evolutionary algorithm based on the surrogate model.Consequently,the optimal control scheme of an actual production optimization problem was obtained,and the Egg model was adopted to test the proposed method via the SADE-PCA method.The results show that the proposed surrogate-assisted evolutionary algorithm based on principal component analysis can effectively reduce the dimension of variables and solve large-scale production optimization of reservoirs efficiently.The optimized injection-production control scheme can effectively increase cumulative oil production and reduce water production to improve the economic benefits of oilfields.In addition,the interpolation control method(ICM)has great performance in achieving smooth control schemes and improving calculation efficiency.