Intelligent Lost Circulation Monitoring with Unbalanced Data Samples

In order to improve the accuracy of intelligent lost circulation monitoring models with unbalanced data samples, a lost circulation monitoring method using the synthetic minority oversampling technique (SMOTE) and bidirectional long short-term memory (Bi-LSTM) network is proposed to efficiently explore the limited and unbalanced drilling data samples. Firstly, the SMOTE algorithm is applied to the field drilling data samples including 10299 data points to produce a new data set that includes 15154 data points. Then, a lost circulation monitoring model based on the Bi-LSTM network is established and trained using the balanced new data set. Experimental results demonstrate that the lost circulation recognition accuracy is improved from 88.93% to 98.38% by the proposed method compared with the widely used LSTM model with unbalanced data samples.