Abstract
The Qiangtang Basin is the most promising new block for oil and gas exploration and development in China. The Jurassic Buqu Formation limestone is a favorable reservoir, but its rock mechanics characteristics are still unclear, and modeling methods for rock mechanics parameters have yet to be determined. Using the Buqu Formation rocks in the Xiangyang Lake area of the Qiangtang Basin as the research object, the mechanical behavior characteristics of Buqu Formation rocks were studied based on laboratory uniaxial compression tests, and the mechanical characteristics of various lithological reservoir rocks of the Buqu Formation were revealed. Based on the experimental results, the applicability of traditional crossplot regression and machine learning modeling methods was compared and analyzed, and an intelligent integrated prediction model framework for rock mechanics parameters was constructed. The results indicate that the Buqu Formation reservoir in the study area has obvious elastic-brittle characteristics, and the rock samples mainly occur splitting failure under uniaxial conditions. Different lithology reservoirs exhibit distinct deformation characteristics and failure modes. The traditional crossplot modeling has a relatively low prediction accuracy (MAPE > 20%). The new machine learning model enhances prediction accuracy and robustness, resulting in high-precision predictions of rock mechanics parameters in the Buqu Formation reservoir (MAPE < 10%) and high consistency with measured data in the verification well (MAPE < 7%). This model can provide a transferable technical foundation for predicting mechanical properties of tiny sample set in underexplored locations.