Assisted History Matching using combined Optimization Methods

Paulo Henrique Ranazzi¹; Marcio Augusto Sampaio Pinto¹

¹ Departamento de Engenharia de Minas e de Petróleo, Escola Politécnica, Universidade de São Paulo

doi:10.20906/CPS/CILAMCE2017-1042

Resumo

The numerical simulation is a fundamental tool for petroleum reservoir management, being the most used to realize the prediction of a field during its productive life. Because the uncertainty parameters used, a discrepancy between the real and simulated values may occur, being necessary the validation of the model, which is made through the history matching (HM). This consists in varying the uncertain parameters iteratively in such a way that the discrepancy between the observed and simulated values is reduced to an acceptable error level. In this work, the methodology of this matching was performed in two steps: re-evaluating 1) the uncertain geological petrophysical properties using random sampling in order to select the best petrophysical images; 2) the productivity index of each well using evolutionary algorithm, since during the construction of the model, loss of information may occur because the process of upscaling. Using the images found in the first step of the methodology, the second step is performed, with the parameter selected to modify the productivity of the wells being the skin factor. The adjust quality was verified using an objective function, which measures the relative difference between the simulation results and observed values. After the matching, the adjusted models were extrapolated to verify their behaviors throughout the field productive life. This methodology was performed in the UNISIM-I-H benchmark model to validate it. The field fluid model was black-oil with the oil density equal to 28 ºAPI, the data consists of 11 years of production history from 15 producers and 10 injectors wells. In conclusion, considering the skin factor as uncertain parameter with the objective of altering the wells behavior resulted in improvements in the matching process. The results showed a reduction in the value of the objective function from 23 to 7 percent.

Palavras-chave: Assisted History Matching; Optimization Methods; Uncertainties