Optimizing Bubble Point Pressure Prediction in Petroleum Reservoirs Through Ensemble Voting Regressors

Abstract Pressure-volume-temperature (PVT) properties are crucial in the oil and gas industry for reservoir modeling, reservoir and fluid characterization. Furthermore, these properties, which include, bubble point pressure, dew point pressure, oil formation volume factor, viscosity and more help determine the behavior of hydrocarbons under different conditions, aiding in efficient extraction and processing. With bubble point pressure as the author's main focus, important practices such as facility design for efficient handling of two-phase production, choice of recovery strategies, dictation of gas liberation from reservoir fluids and optimization of production rates, all have this property as their determining factor. However, accurately predicting the bubble point pressure in the oil and gas industry poses a significant challenge, given the time-consuming, expensive, and often inaccurate nature of existing methods like the empirical and experimental approaches. Hence, the purpose of this paper is to present an intelligent system approach of developing an ensemble voting Regressor model for the prediction of bubble point pressure. Trained with 604 data points, from oil fields all over the world, with the input parameters like API oil gravity, gas specific gravity, reservoir temperature, gas-oil ratio, the hybrid model was found to accurately predict bubble point pressure. Also, a comparative analysis, showed that the model outperformed pre-existing correlations with a 92% accuracy. The results of this study help better the understanding of the behavior of hydrocarbon reservoir fluids and further optimizes petroleum extraction processes.