Análise do comportamento de fluidos de reservatório com diferentes teores de co2 e arranjos de pseudocomponentes

Abstract

A highly complex task consists in the simulation of reservoirs containing a complex mixture of hydrocarbons with high CO2 content, associating, simultaneously, the precision, speed and economy in computational processes using the current resources. Therefore, to solve such problems, it is necessary to make decisions based on the use of compositional simulation, simplifying and modeling fluid properties that are acceptable for the reproduction of the physical behavior of the system and capable of reducing the computational time to an acceptable threshold. The way in which the lumping and the regression of the parameters are performed for different CO2 concentrations, besides the choice of the equations of state, are critical and determinant in the reliability of the compositional simulations. Therefore, the aim of this work was to conduct a study of how the representation of petroleum fractions by pseudocomponents and choice of state equation models are able to change the description of the behavior of the oil sample, submitted to its injection in a reservoir with high pressure. In this analysis, in order to analyze the influence of the lumping process, five arrangements (5, 7, 9, 15 and 24 pseudocomponents) were compared using an oil found in the literature with a content of 3.55% CO2. For each arrangement, two state equations were applied and compared: Peng Robinson (PR) and Soave Redlich Kwong (SRK) equations. In addition, to evaluate the behavior of the algorithm in mixtures with higher CO2 content, a fluid with 8.24% content was analyzed. Differential liberation and saturation pressure tests were used, in which the pseudizations for both cases were automated with the aid of an algorithm. Therefore, with the composition of 3.55% CO2, the curves presented variations in the lumping arrangements and changes in the equations of state. Regarding the fluid with 8.24% CO2, the results generated from the tests used in the study were satisfactory, showing the good performance of the algorithm in both cases.