Reliable estimation of organic matter characteristics is essential in drilling decisions, source rock evaluation, and unconventional reservoir production. Their measurement is based on experiments after core sampling, which is time-consuming and economically challenging. In this study, we present a new approach to evaluate the characteristics of organic matter in source and reservoir rocks by in-situ electrical heating and temperature transient analysis under in-situ conditions.
The new approach is based on inverse modeling, which monitors in-situ heater temperature during electrical heating and machine learning technologies. Thermal method of electrical heating is applied for the in-situ pyrolysis, to figure out the characteristics of organic matter—kerogen volume fraction and activation energy of decomposition reaction. The heater temperature acts as an indicator of type and maturity of kerogen, since it is affected by the bulk thermal conductivity of formation, which is a function of dynamically changing rock-and-pore composition by kerogen decomposition. A full-physics simulation model of in-situ kerogen pyrolysis is used to generate output data of electrical heater temperature, which is the input data of learning-based models. Minimal simplification of physical and chemical phenomena in the full-physics simulation model, which describes the multicomponent-multiphase-nonisothermal systems involving kinetic reactions, gives the confidence of synthetic output data of heater temperature.
Full-physics simulation model computes system responses under unknown and uncertain input parameters, which determine the reactivity of kerogen pyrolysis. The full-physics simulation model generates the sets of heater temperature transient data while heating with constant heat flux, in the 300 different simulated source rocks containing Types 1, 2, and 3 kerogens with various organic matter content and activation energies. Based on the set of heater temperature transient data as input parameters, Artificial Neural Network (ANN) is employed to generate a black box model to estimate the unknown organic matter content and activation energy. Developed ANN data-driven model shows better performance in estimating unknown parameters, in Types 2 and 3 kerogens with wide ranges of activation energies than Type 1 kerogen with a narrow range of activation energy. Support Vector Machines (SVM) method, which categorizes data into multiple classes by using hyperplanes, is applied to classify the heater temperature transient data into different types of kerogens and shows good performance in classification.
The new characterization technology of in-situ organic matter in source rocks presented in this study provides reliable information of types and maturity of organic matter, without experiments after core sampling. It is expected to enable the realistic evaluation of source rocks under subsurface conditions, by resolving technical and economic challenges.