Prediction of petroleum formation: the influence of laboratory heating rates on kinetic parameters and geological extrapolations

Abstract

While major variations in predictions of petroleum formation timing are supposed to be related to the burial history of the source rock and the type of organic matter preserved therein, it is often ignored that the laboratory techniques used to obtain reliable kinetic parameters for the description of petroleum formation and its extrapolation to natural conditions can be problematic. In the present paper bulk petroleum generation by open-system programmed-temperature pyrolysis of Green River Shale, Toarcian Shale and Westfalian coal samples has been studied comparatively at heating rates of 0.1, 0.7, 5.0, 15.0 and 25.0 K/min in order to evaluate the influence of laboratory heating rates on kinetic parameters and geological extrapolations. Kinetic parameters were calculated from three sets of pyrolytic generation rate versus temperature curves measured at either 0.1, 0.7 and 5.0, 0.7, 5.0 and 15.0 or 5.0, 15.0 and 25.0 K/min using both the distributed and the T max-shift model. Frequency factors and dominant activation energies derived from the T max-shift model are found to decrease systematically in going from sets of slow to fast heating rates. For Type II and III kerogens this decrease of kinetic parameters is confirmed by the distributed model; for algal kerogens of Type I the distributed model cannot be applied. The decrease of kinetic parameters results from increasing delay in heat transfer between the inert gas stream and the sample with increasing rate of heating. Reducing the sample weight can partly but not fully compensate this effect. By using the natural maturation series of Westphalian Coals and Toarcian Shales it was possible to illustrate that geological extrapolations from slow heating rate experiments fit much better to natural petroleum formation than those from fast heating rates. The latter lead to temperature predictions which are too low by 30–40 °C, which in turn would result in a 1000 m failure when predicting the oil window depth in an average sedimentary basins.