Abstract
With the expansion of oil and gas exploration and the development of complex oil and gas resource areas, the probability of risk is increasing. Kick is one of the high-risk risks of drilling, and timely and accurate early warning is increasingly important. Based on the kick generation mechanism, kick characterization parameters are preliminarily selected. According to the characteristics of the data and previous research progress, Random Forest (RF), Support Vector Machine (SVM), Fully Connected Neural Network (FNN), and Long Short-term Memory Neural Network (LSTM) are established using experimental data from Memorial University of Newfoundland. The test results show that the accuracy of the SVM-linear model is 0.952, and the missing alarm, the false alarm rate is only 0.064, 0.035. Also, through the analysis of the kick response time, the lag time of the SVM-linear model is 3.6s, and the comprehensive equivalent time is 23.13, which shows the best performance. This paper lays a good foundation for the establishment of an intelligent kick early warning model.