A Hybrid Neural Network Model for Predicting Bottomhole Pressure in Managed Pressure Drilling

Managed pressure drilling (MPD) is an essential technology for safe and efficient in deep high-temperature high-pressure formations with narrow safety windows. However, the complex conditions wells make mechanism of multiphase flow annulus complicated increase difficulty accurate prediction bottomhole (BHP). Recently, increasing volume research shows that intelligent means accurately predicting BHP. few studies have focused on temporal properties variation In this paper, hybrid neural network models based multi-branch parallel are established by combining different advantages back propagation (BP), long short-term memory (LSTM), a one-dimensional convolutional (1DCNN) model. The results show relative error best model about 70% lower than optimal single Preliminary experimental reveal combine models, which more robust extracting features MWD. Finally, trend analysis, validity further verified. This study provides reference solving problem optimizing characteristics guidance fine control formations.