Abstract

Abstract Oil and gas drilling is a complex construction process, drilling conditions as the key parameters of the construction process, efficient and accurate identification of drilling conditions is the basis for statistical drilling efficiency and analysis of drilling status. With the continuous development of integrated logging technology and sensor technology, field operators and researchers have access to large amounts of realtime data. The existing methods mainly include logical judgment and manual judgment, which have the problems of insufficient accuracy and low efficiency, respectively. In recent years, more and more scholars have adopted data-driven methods such as machine learning to identify drilling conditions, but pure data-driven models have the problem of high false positive rate. This paper proposes an intelligent identification method for drilling conditions based on a combination of deep learning and drilling experience knowledge. First, abnormal fluctuations in the logging data are removed through an efficient real-time data cleaning process, and then the most critical parameters that characterize the drilling conditions during the drilling process are selected. Three deep learning algorithms, Gate Recurrent Unit (GRU), Long short-term memory (LSTM), and Convolutional Neural Networks (CNN), are used to build the intelligent identification model for drilling conditions and compare them. Then, according to the expert knowledge of drilling condition judgment, the logical constraint is formed, and it is added to the identification process. The final workflow can effectively realize real-time intelligent identification of 10 drilling conditions. The recognition precision and recall rate of various conditions are both above 94%, providing a basis for drilling efficiency analysis and risk prevention.