Intelligent Method for Lithology Determination in Drill-following Formations based on GA-GRNN Model
He Zhang, Jinhua Yang · Frontiers in Sustainable Development · 2025
In oil and gas drilling operations, formation rock lithology identification is of great significance to the whole oil and gas drilling operation process. Aiming at the problems of poor real-time performance and insufficient model accuracy caused by the poor data quality of traditional lithology identification methods in the drilling process, this study proposes an intelligent identification method based on the optimization of generalized regression neural network (GRNN) by genetic algorithm (GA), which firstly performs the dimensionality reduction of the logging element data by correlation analysis and principal component analysis, and then searches the global optimal values of the smoothing parameters of generalized regression neural network (GRNN) by genetic algorithm (GA) to train the optimal identification model. The experimental results show that the recognition accuracy of the GA-GRNN model is improved by 12.1% and 7.5% to 92.9% compared with the BP neural network and standard GRNN model, respectively., and the field application verifies the engineering applicability of this method in real-time lithology recognition in X oilfield with the accuracy of 94% in the recognition of complex formations, which provides a new technological means for intelligent drilling decision-making.