Assessment of Different PEM Calibrations Across Multiple Datasets for 4DQI

Petroelastic models (PEM) play a crucial role in 4D quantitative interpretation. The lithological variation of a PEM is captured by rock frame models which are a function of mineral elastic moduli and hyperparameters. Assigning a single hyperparameter for the whole reservoir and universal values for different reservoirs may lead to erroneous 4D responses. In this study, the applicability of calibrating seven rock frame models on some data sets using a grid-search algorithm and a probabilistic approach for optimisation is evaluated. The results show that the empirical models could model the observed elastic moduli of clastic and carbonate rocks better than inclusion-based and contact-based models. In addition, although it is widely expected that contact-driven models are more representative of rock physics characteristics of clastic rocks, inclusion-based models show better performance in modelling elastic moduli of such reservoirs. Moreover, it is observed that porosity is linearly correlated with the optimised depth-dependent critical porosity and pore aspect ratio, while both porosity and shale volume are linearly correlated with the coordination number. The results of calibrating the Xu and Payne model confirm the performance of the proposed probabilistic approach to calibrate complex models.