Data-Driven Method for Predicting the Transportable Maximum Gas–Oil Ratio

Abstract

In the process of oil–gas mixing transportation, a too-high gas–oil ratio (GOR) will lead to instable flow pattern and high pipeline pressure, which has great safety risks. Therefore, it is important to determine the maximum GOR. At present, it relies mainly on commercial software to simulate the operation of the mixture transportation pipeline and the hydrothermal operating parameters to determine the maximum GOR under certain condition. This enumeration method is time-consuming and does not apply to continuously parameters. To solve this problem, a data-driven predictive model is developed. The new features are constructed by analyzing the factors influencing the transportable maximum gas–oil ratio (TMGOR), and the highly correlated features are selected from them as the new features. After analyzing the characteristics of the target variables, data mapping is performed, and the processed data set is fed into a neural network for training to obtain a data-driven predictive model of TMGOR. Finally, the validation is carried out with field data from an oilfield block in northwest China. The results showed that the average relative error of the model does not exceed 8.2% compared with the simulation results of commercial software, which has a high accuracy and can provide a rationale for the decision-making of mixed transfer in the field.