Intelligent Prediction Method and System for Clay Mineral Transformation Stage in Continental Shale Oil Reservoirs

Abstract

This invention provides an intelligent prediction method and system for clay mineral transformation stages in continental shale oil reservoirs, belonging to the field of clay mineral transformation stage prediction. To address the problems of existing methods for judging clay mineral transformation stages being highly dependent on experts and having high sampling costs, this invention includes: establishing a geological parameter dataset covering Ro and multi-dimensional logging curve data; completing clay mineral transformation stage calibration through core-logging response coupling annotation; constructing regional geological feature maps using the KNN clustering algorithm; designing a temporal convolutional network incorporating a multi-head attention mechanism for areas without thin section and SEM sample data, extracting local abrupt change features of logging time-series signals through parallel multi-scale convolutional kernels to achieve dynamic modeling with minimal human intervention; constructing a dual-channel joint prediction model, simultaneously performing graph spatial convolution and temporal feature decoding, optimizing cross-modal weight allocation, and outputting a comprehensive prediction result of the mineral transformation stage that integrates spatiotemporal evolution laws.

Images