Machine learning-based prediction of well performance parameters for wellhead choke flow optimization.

Area of Science:

• Petroleum Engineering
• Machine Learning Applications
• Fluid Dynamics

Background:

• Accurate fluid flow rate forecasting is crucial for hydrocarbon recovery and production optimization.
• Wellhead chokes are vital for stable downstream pressure and controlling in-well pressure drops.
• Existing multiphase flow models have limitations in global applicability and accuracy.

Purpose of the Study:

• To evaluate machine learning algorithms for predicting well performance parameters.
• To compare the predictive accuracy of Convolutional Neural Network (CNN), Multilayer Perceptron (MLP), and Radial Basis Function Network (RBFN).

Main Methods:

• Employed three machine learning algorithms: CNN, MLP, and RBFN.
• Utilized a dataset with five input parameters: liquid production rate, wellhead pressure, choke size, BS&W, and GLR.
• Evaluated models using R-squared, RMSE, MSE, MAPE, and MAE metrics.

Main Results:

• MLP demonstrated the highest predictive performance with R-squared values up to 0.9985.
• MLP achieved low Root Mean Square Error (RMSE) values of 0.0024 (training) and 0.0057 (testing).
• The dataset was split into 70:30 for training and testing.

Conclusions:

• MLP is a highly effective model for predicting well performance parameters in hydrocarbon production.
• Machine learning offers a more accurate alternative to traditional models for well flow rate forecasting.
• The study highlights the potential of AI in optimizing oil and gas production.