Real-Time Liquid Rate and Water Cut Prediction From the Electrical Submersible Pump Sensors Data Using...

Area of Science:

• Petroleum Engineering
• Data Science
• Artificial Intelligence

Background:

• Traditional multiphase flow rate calculations in oil and gas wells are costly and lack trend identification.
• Virtual Flow Metering (VFM) is gaining traction as an efficient alternative in the industry.
• Electrical submersible pumped (ESP) wells present unique challenges for accurate flow rate estimation.

Purpose of the Study:

• To develop and validate a novel data-driven approach for real-time multiphase flow rate calculation in ESP wells.
• To present a cost-effective and accurate alternative to physical multiphase flow meters and production testing.
• To demonstrate the efficacy of integrated machine learning models for VFM in ESP applications.

Main Methods:

• A robust workflow combining symbolic regression, extreme gradient boosted trees, and a deep learning model (CNN-LSTM).
• Utilizing data from pump sensors to predict liquid rate and water cut in real time.
• Developing a pipeline of convolutional neural network (CNN) and long short-term memory (LSTM) layers for predictive modeling.

Main Results:

• The proposed data-driven VFM approach accurately predicts liquid rate and water cut in real time.
• The methodology proves to be a cost-effective solution compared to conventional flow metering techniques.
• The study highlights the significant potential of data-driven methods for VFM in ESP wells.

Conclusions:

• The developed VFM method offers a reliable and economical solution for monitoring and optimizing production in ESP wells.
• This data-driven approach enhances the capabilities of virtual flow metering in the oil and gas sector.
• The findings underscore the effectiveness of combining advanced machine learning techniques for real-time well performance analysis.