Optimization of oil field development based on a 3D reservoir model obtained as a result of history matching

The paper proposes an approach to optimizing the development of oil fields. The objective function includes weighted squares of development target indicators and regularizing terms, in which the coefficients are searched adaptively. Regularizing terms ensure the fulfillment of restrictions on the optimized parameters and the rapid convergence of the optimization process. When minimizing the objective function, linearization of the target indicators is performed, and the values of the optimized parameters at the next iteration are sought by solving the system of linear algebraic equations obtained from minimizing the quadratic functional. The values of the target indicators and their sensitivity to the parameters being optimized are calculated by fluid dynamic 3D modeling for the oil reservoir model obtained as a result of history matching for the period preceding the optimization period. Calculations are performed in a distributed computing system consisting of multi-core personal computers. To test the proposed approach, a model of a high-viscosity oil field in Tatarstan was used. The optimization was carried out with various weighting factors and desired oil recovery values in the corresponding target indicator. It is shown that the optimized plans provide more efficient development of the oil field compared to the plan used in practice. At the same time, the optimal plan, built on the basis of a reservoir model history-matched at an early stage of development, optimizes development for a model history-matched throughout the entire period of field development. This allows us to conclude that development plans obtained from a model history-matched using a short time period will optimize production characteristics for a real field to about the same extent. The time for solving optimization problems containing about 500 parameters in a distributed computing system was about a day.