Machine learning-based fracturing parameter optimization for horizontal wells in Panke field shale oil.

Area of Science:

• Petroleum Engineering
• Machine Learning Applications
• Reservoir Optimization

Background:

• Multistage fractured horizontal wells (NFHWs) are crucial for maximizing production in tight oil and gas reservoirs.
• Optimal fracturing parameter design is essential for enhancing well productivity and economic returns.
• Accurate production prediction is vital for effective development strategies and planning.

Purpose of the Study:

• To develop a novel workflow integrating machine learning (ML) and particle swarm optimization (PSO) for production prediction and fracturing parameter optimization in tight reservoirs.
• To establish a data-driven approach for maximizing net present value (NPV) and optimizing fracture designs for NFHWs.
• To overcome limitations in current forecasting methods that struggle to balance economic returns with operator preferences.

Main Methods:

• Conducted 10,000 numerical simulation experiments to generate a comprehensive dataset for training and validation.
• Developed and evaluated five ML models for production prediction, selecting Random Forest (RF) as the best performer.
• Integrated the validated RF model with PSO to optimize fracturing parameters for maximizing NPV, creating the ML-PSO hybrid model.

Main Results:

• The Random Forest model demonstrated superior performance in mapping feature variables to production yield.
• The hybrid ML-PSO model effectively optimizes fracturing parameters to maximize net present value.
• The workflow provides accurate, real-time production prediction and serves as a tool for optimizing fracture design.

Conclusions:

• The proposed ML-PSO workflow offers an efficient and accurate method for predicting production and optimizing fracturing parameters for NFHWs in tight reservoirs.
• This approach facilitates data-driven decision-making, improving economic outcomes in oil and gas development.
• The study provides a foundation for advanced forecasting models in the oil and gas industry, particularly for tight formations.