Deep Application of Artificial Intelligence in Predicting Geological Sweet Spots in Low-Permeability Terrestrial Oil Fields

Abstract:

To address the limitations of traditional geophysical methods in effectively predicting the distribution of low-permeability sweet spots in deep continental formations， a high-precision geological sweet spots prediction method based on artificial intelligence （AI） is proposed， providing reliable guidance for optimizing fracturing schemes and deploying well locations. This multi-step AI-driven approach incorporates XGBoost algorithm for constructing an improved empirical quality index （RQI） for low-permeability reservoirs， bee colony algorithm deep neural network for 1D well-point RQI curve prediction， and improved CNN algorithm based on Dropout strategy for spatial three-dimensional RQI distribution prediction.This methodology breaks through the limitations of traditional sweet spots prediction techniques， achieving multi-scale intelligent prediction from 1D to 3D while significantly improving the accuracy of low-permeability reservoir characterization. The application in a certain oil field shows that this method effectively overcomes the influence of various unfavorable factors and significantly improves the accuracy of geological sweet spots prediction compared to traditional methods.