Using Artificial Neural Network To Calculate Pressure Drop in Multiphase Flow

Abstract

It is immense important to have accurate prediction of pressure drop over the life of the well to design more effective tubing and optimum production operation. A deployment of pressure gauge is a common practice to measure flowing bottom-hole pressure (FBHP). Additionally, some mechanic models and empirical correlations for multiphase have been proposed in order to avoid a significant expense and time-consuming of intervening a producing well. Nevertheless, the recent prediction techniques present a low level of accuracy in the result, the improvement method is needed to tackle this problem. This paper reports our recent study on the use of an Artificial Neural Networks (ANN) to predict the pressure drop in multiphase flowing wells. The ANN model is developed based on numerous different surface production data including oil flow rate, gas flow rate, gas-oil ratio, wellhead pressure, wellhead temperature, bottomhole pressure, bottomhole temperature. The collected data sets from the flowing well X at Hai Thach Moc Tinh field are normalized and imported into the ANN models. The proposed models covered a wide range of variables with different neuron number of hidden layers. The results between different data set are records and compared statistically to each other in order to choose the least error.