Confidence limits on kinetic models of primary cracking and implications for the modelling of hydrocarbon generation

Abstract

Kinetic models of primary cracking for use in hydrocarbon modelling are obtained through empirical calibration against laboratory measurements. This paper utilizes least-squares variance analysis to evaluate the uncertainty in the calibrated kinetic parameters and discusses some consequences for the prediction of hydrocarbon generation. It is substantiated that the sensitivity of the average kinetic parameters E a ∗ and log e (A) ∗ to micropyrolysis errors can be parameterized by errors in pyrolysis T max. The absolute and relative errors in T max are modelled by normal (Gaussian) random variables and the variance analysis yields the joint normal probability density distribution of variability in E a ∗ and log e (A) ∗ . It is found that T max errors reduce the ability to resolve independently the values of E a ∗ and log e (A) ∗ : the estimation errors are highly positively correlated and repeated measurements on identical kerogens will produce values of E a ∗ and log e (A) ∗ which tend to fall close to and along the line log e (A) ∗ = E a IRT p + constant , where T p is an average pyrolysis temperature. This correlation is consistent with observations, and hence it is suggested that it is mainly an artifact which originates in the experimental errors of laboratory pyrolysis. When values of E a ∗ and log e (A) ∗ are confined to the region of the confidence ellipsoid, the pyrolysis data are satisfied within their error bounds. However, the variability in predictions of hydrocarbon generation may be large depending on the particular geological temperature history. The uncertainties in the depth to the oil window may be several hundreds of metres, and the timing of peak hydrocarbon generation rate may vary considerably. The inclusion of basin geochemical data in the calibration drastically reduces the ambiguity in E a ∗ and log e (A) ∗ and emphasizes the need to validate the laboratory calibrated kinetic models against basin observations.