Research on Drilling Overflow Feature Extraction and Data Processing Method based on Real-time Data

In this study, an overflow early warning method based on logging big data and machine learning algorithm was proposed to solve the problem of early overflow monitoring. By analyzing the overflow generation mechanism, the real-time drilling dataset containing 25 original features was preprocessed by feature engineering technology, and then the 12 initial overflow-related parameters were optimized and screened by recursive feature elimination (RFE), and finally 9 core features were selected as model inputs. The results show that data-driven feature engineering can effectively improve the generalization ability of the overflow early warning model, gain valuable time for well control response, and have important practical value for reducing the risk of drilling accidents under complex geological conditions.