Prompt assessment of the economic potential of field at the early stage of development by population-based optimization algorithms

Relevance. The necessity of operative selection of the optimal development system, one of the key parameters of which is the location of the well stock. At present deterministic approaches based on a limited number of geological realizations (percentiles of 10, 50, 90 or profitability criteria) followed by analytical evaluation of well production profiles are used for economic evaluation of the object potential. These approaches do not allow fully evaluating the system and making reasonable decisions when planning the field development system under conditions of high geological variability and uncertainty. It is possible to increase the number of estimated geologic realizations by upscaling the simulation model. It is assumed that the proposed approach with population optimization algorithms will allow automatic selection of the stock under geological uncertainty in order to probabilistically assess the economic potential of the object and reduce costs while designing field development options. Aim. Improvement of approaches for operational selection of parameters of optimal location and length of producing wells based on optimization algorithms to assess the economic potential of the development object and justify the optimal design solution. Object. Reservoir of oil and gas condensate field in Eastern Siberia. Methods. Mathematical modelling, optimization algorithms and statistical methods. Results. The proposed approach includes upscaling of the model and usage of population optimization algorithms to eliminate the disadvantages of the conventional approach. The selected approach allowed finding the optimal well stock with minimal error of volumes of produced target fluid compared to full-scale precise 3D model, which also provided high convergence of economic calculations. The particle swarm optimization method provided comparable results with a smaller number of iterations compared to the genetic algorithm, but allowed for shorter optimization times. The discrepancy between the models for the main indicators remained within acceptable values (oil up to 12%, net present value less than 5%). The average optimization cycle took 400 iterations and 130 minutes. The proposed approach with upscaling of the simulation model to find the optimal well stock reduced the calculation time by 10 times.