Improved Characterization of Non-Linear Injectivity Formation Damage from Coreflood Pressure Measurements

Abstract

Abstract PWRI and seawater injection may result in drastic injectivity decline. The reason is solid and oily particles, which are captured by rock from injected water causing steep permeability decline. The reliable injectivity decline prediction, allowing for injector stimulation planning and design, is based on mathematical modelling with well-known values of two model functions of filtration and formation damage. Previous works proposed the method to determine two injectivity damage constants - filtration and formation damage coefficients - by measuring pressure drops on the core and also on its first section (so called 3-point-pressure method). Dimensionless pressure drop, which is proportional to skin factor, was proven to be linear with time, allowing for unique solution of the laboratory data matching problem. Yet, often the skin growth curves are non-linear. It is explained by the retained concentration dependence of filtration and formation damage functions. In the previous works, the physics reasons for filtration functions to be linear have been analysed. Therefore, the filtration coefficient was assumed to be linear function of retained concentration. The inverse to the formation damage function was assumed to be quadratic polynomial of retained concentration, which caused non-realistic skin prediction and non-unicity of the matching problem. In the present paper, the inverse to the formation damage function was assumed to be monotonic power function of retained concentration, allowing for unique matching with laboratory data. An analytical model developed accurately describes pressure data of several coreflooding tests. In order to validate the proposed method, the pressure drop was measured on the second core section on top of the first section and the overall core. The measured data were compared with modelling tuned by the data from the first section and the overall core. Good agreement between the measured and fully predictive modelled data was observed, which validates the proposed method.