The application of back propagation neural network for predicting production flow rate of oil wells in Hai Su Trang field, Vietnam

Oil production flow rate prediction is a critical aspect of oil and gas exploitation operations. Currently, flow rate forecasting is often estimated using theoretical or empirical models. Theoretical models tend to provide predictions with a wide range of errors and require extensive input data. On the other hand, empirical models have limitations due to restricted data. The objective of this article is to develop a correlation for highly accurate forecasting of the production flow rate. In order to achieve this goal, this study applies an artificial neural network (ANN) for flow rate prediction. The backpropagation algorithm and the tansig function are selected in this study as a learning algorithm to forecast flow rate. The study considered 262 datasets collected from six wells in the Hai Su Trang field, Cuu Long basin used in the ANN model, with 70% for training, 15% for testing, and the remaining 15% for validation. This article evaluates the ability of ANN model to predict flow rate with different numbers of neuron. The predicted results obtained from the ANN model with eight neurons and backpropagation algorithm achieved high predictability when compared to empirical methods and multivariate regression model, with a strong correlation coefficient of 0.97 and a low RMSE of 32.54 bbl/d. Therefore, the developed ANN models have been shown to be an effective tool in production flow rate forecasting in oilfields.