Real-time warning method for sand plugging in offshore fracturing wells

Area of Science:

• Petroleum Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Sand plugging is a major cause of hydraulic fracturing failures.
• Current detection methods are slow, inaccurate, and lack real-time capabilities.

Purpose of the Study:

• To develop an intelligent method for real-time sand plugging identification and warning.
• To improve the accuracy and efficiency of sand plugging detection in hydraulic fracturing.

Main Methods:

• Utilized offshore hydraulic fracturing data.
• Employed Attention Mechanism based Long-Short Term Memory Network (Att-LSTM) for real-time pressure prediction.
• Developed an Attention Mechanism based Convolutional Long-Short Term Memory Network (Att-CNN-LSTM) for sand plugging identification.
• Integrated Att-LSTM and Att-CNN-LSTM with transfer learning for continuous learning.

Main Results:

• Att-LSTM achieved >92% accuracy in pressure forecasting within 40 seconds.
• Att-CNN-LSTM identified sand plugging with <1 minute error margin.
• The integrated system significantly improved accuracy and efficiency in sand plugging detection.

Conclusions:

• The proposed intelligent system offers a novel and effective solution for real-time sand plugging warning.
• This approach enhances safety and efficiency in hydraulic fracturing operations.
• The study provides critical technical support for intelligent decision-making in fracturing.