无梯度多参数自动历史拟合方法

摘要

The uncertainties of the reservoir geological parameters can be reduced by history matching. A stable and efficient multi-parameter optimal adjustment technology was proposed in this work. This technology involved setting up a mathematical minimization model of history matching based on Bayesian statistical theory and reducing the dimension of parameters by sin-gular value decomposition method. And the optimization problem was finally solved by using gradient-free method based on directional derivative. By using this method, permeability, porosity, relative permeability curve, viscosity of crude oil and the oil-water interface parameters can be matched at the same time to further reduce the uncertainties of the inverse problem. Compared with manual history matching, automatic history matching results based on dimension reduction method show a high degree of consistency which proves the effectiveness and the correctness of this method. Automatic history matching process is also proved to be capable of saving a lot of human and machine labor.%通过历史拟合以降低油藏地质参数的不确定性，提出一种稳定高效的无梯度多参数最优化调参技术，根据贝叶斯统计理论建立历史拟合最小化数学模型，并利用奇异值分解算法进行降维分解，然后采用基于方向导数的无梯度算法对该最优化问题进行求解。该方法同时实现了渗透率、孔隙度、相对渗透率、原油黏度及水体参数的拟合，进一步降低了反问题的不确定性。与人工历史拟合结果相比，基于降维方法的自动历史拟合结果表现出高度的一致性，节省了大量的人力与机时，证明了该方法的有效性与正确性。