Determination of Basic Reservoir Parameters in Shale Formations as a Solution of Inverse Problem in the Computer Assisted History Matching of their Simulation Models. Part II – Hybrid Optimization Algorithm

Piotr Łętkowski, Wiesław Szott
2016 Nafta - gaz : miesiecznik poswiecony nauce i technice w przemysle  
The paper presents a hybrid optimization algorithm as a practical method to solve the inverse problem of simulation model calibration process. The method is applied to determine basic reservoir parameters in shale formations as a result of the process carried out for their models. Due to some specific features of the problem including its nonlinearity and the large size of the solution space, an algorithm that can be employed in the process of model automatic calibration has to run fast and be
more » ... ffective in finding acceptable solution using limited number of simulations. The selection of an appropriate global optimization method is crucial in the situation of many expected local minima of the problem. One of the stochastic sampling method used and presented in the paper is the method of Particle Swamp Optimization (PSO). Such a method is characterized by a simple concept, fast convergence, and intelligent balance between searching and testing of the solution space. Besides the PSO method three other elements are combined to result in the effective solution of the problem. They include: search with stable Levy distribution of iteration step size, Latin hypercube sampling and response surface. The combination of the elements employs both deterministic and stochastic approaches that make the proposed solution both robust and effective. The algorithm was positively tested for convergence and performance using special functions that are commonly applied for such purposes. Key words: reservoir simulation models, inverse problem, optimization methods, particle swarm optimization. niki deterministyczne i stochastyczne, co pozwala na wyeliminowanie wad każdej z metod. Ponadto przedstawiono wyniki testów zbieżności zbudowanego algorytmu, potwierdzając przy tym jego efektywność przy przeszukiwaniu przestrzeni rozwiązań. Słowa kluczowe: symulacyjne modele złożowe, problem odwrotny, metody optymalizacyjne, optymalizacja rojem cząstek.
doi:10.18668/ng.2016.10.06 fatcat:id5m2asnmbfu7gqcinvnfb7vgu