Prediction of the stability and compatibility of Colombian heavy crude oils by 1D low field nuclear magnetic resonance relaxometry and chemometric methods

Abstract

Stability and compatibility can be described by the insolubility number (IN), the solubility blending number (SBN) and the ratio (R) between them, which has become a critical index in daily refineries operation. In this work, a new methodology for predicting the stability and compatibility of heavy crude oils by 1D low field nuclear magnetic resonance (LF-NMR) relaxometry has been developed. Various chemometric models were developed to predict the IN,SBN and R properties of Colombian heavy crude oils (°API from 6 to 27), by integrating intervals of either the 1D free induction transverse magnetization decay curves (T2 FID) obtained with a low-field (LF) nuclear magnetic resonance (NMR) spectrometer, or their T2 relaxation distributions obtained via the inverse Laplace transform (ILT). Correlations between the relaxometry curves and said properties were evaluated using the principal component regression (PCR) and the partial least squares regression (PLSR) methodologies. The obtained prediction models yielded coefficients of determination (R2) above 95% with as few as 6 components and values of cross validated root mean squared error of prediction (RMSEP) as low as 0.71 for the IN, 2.37 for the SBN and of 0.03 for the R property, respectively. The best prediction models were attained using PLSR and employing the FIDs as the predictors. The use of chemometric predictive models from LF-NMR measurements represent a low-cost and faster alternative to estimate critical properties of the independent crude oils to be blended, and thus, will help in predicting the stability and compatibility of crude oil blends, which in turn may help accelerate the decision-making process in daily refineries tasks.