Bottom-Hole Pressure Estimation from Wellhead Data Using Artificial Neural Network

Abstract

Accurate pressure losses prediction for flow in tubing installations is of great importance in the petroleum industry. Historically, the Bottom-Hole Pressure (BHP) determination was obtained using down-hole pressure gauges, because of the economic disadvantage and redundancy, this procedure seems to be less effective, this led to the adoption of the BHP prediction process in estimation. The wide acceptance of data-driven analytics makes the estimation procedure a valid approach in the industry today. Recently, the Artificial Neural Network (ANN), a technique which has been widely accepted because the model proved to predict better than the conventional correlations. This present work aims to develop a prediction model for BHP based on input and output data obtained from the Volve production field in Norway. Machine learning algorithm based on ANN was used to predict and further improve the accuracy of the prediction while considering a large production dataset from different wells of the field. In developing the model, the initial dataset was processed to about 2,500 data points; the model was trained, tested and cross-validated based on the parameters from the data. Results affirmed that ANN has the ability to handle large dataset, results also revealed that ANN outperformed other models, with a Coefficient of Determination of 0.99997, Root Mean Squared Error of 0.07405 and Mean Absolute Error of 0.02657, which shows high predictability of the model. These results indicated that the ANN model gives a better prediction of BHP when compared to other mechanistic models. Finally, this work supports the claim, that Production engineers can accurately predict the pressure at the sand-face of a producing well without the use of expensive BHP gauge.