| Popis: |
ABSTRACT: Mud loss is one of the most common and troublesome wellbore problems. Predictive evaluation of mud loss types not only optimizes drilling design, but also reduces potential costs before drilling. To solve mud loss problem of M formation in the H oil field, we proposed a practical solution based on machine learning in this paper, which can predict the mud loss types using seismic data. Firstly, we calculated and obtained 16 seismic attributes in 6 categories, and according to the mud loss rate and volume, we classified the mud loss into 4 types: seepage loss, partial loss, severe loss, and total loss. Then 10 typical wells were selected from 50 wells, which covered different mud loss types and depth. The seismic attributes of single well with the above characteristics were extracted, and the relationship between seismic attributes and mud loss type were obtained using machine learning. Finally, a 3D probability prediction model of potential mud loss type is obtained and analyzed with a practical case. Our model can predict the distribution of mud loss types at different depths in different regions. It can not only be used in the design of well location and well trajectory but also provide scientific suggestions for mud loss prevention and plugging. 1. INTRODUCTION Mud loss is one of the common downhole problems during drilling and completion. It not only causes a significant increase in non-productive time (NPT), but also increases well control risk (Sun et al., 2021). When drilling in the carbonate formation, mud loss has become the primary problem affecting safe and efficient drilling due to the complex and changeable leakage zones (Savari and Whitfill., 2019; Huang et al., 2020). Therefore, mud loss prediction of carbonate formation has become a hotspot and "Difficult topic" in current research. |
