An Accurate Reservoir’s Bubble Point Pressure Correlation

Abstract

Bubble point pressure ( P b ) is essential for determining petroleum production, simulation, and reservoir characterization calculations. The P b can be measured from the pressure–volume–temperature (PVT) experiments. Nonetheless, the PVT measurements have limitations, such as being costly and time-consuming. Therefore, some studies used alternative methods, namely, empirical correlations and machine learning techniques, to obtain the P b . However, the previously published methods have restrictions like accuracy, and some use specific data to build their models. In addition, most of the previously published models have not shown the proper relationships between the features and targets to indicate the correct physical behavior. Therefore, this study develops an accurate and robust correlation to obtain the P b applying the Group Method of Data Handling (GMDH). The GMDH combines neural networks and statistical methods that generate relationships among the feature and target parameters. A total of 760 global datasets were used to develop the GMDH model. The GMDH model is verified using trend analysis and indicates that the GMDH model follows all input parameters’ exact physical behavior. In addition, different statistical analyses were conducted to investigate the GMDH and the published models’ robustness. The GMDH model follows the correct trend for four input parameters (gas solubility, gas specific gravity, oil specific gravity, and reservoir temperature). The GMDH correlation has the lowest average percent relative error, root mean square error, and standard deviation of 8.51%, 12.70, and 0.09, respectively, and the highest correlation coefficient of 0.9883 compared to published models. The different statistical analyses indicated that the GMDH is the first rank model to accurately and robustly predict the P b .

Field Telemetry Drilling Dataset Modeling with Multivariable Regression, Group Method Data Handling, Artificial Neural Network, and the Proposed Group-Method-Data-Handling-Featured Artificial Neural Network

“…To study these effects, the GMDH regression was considered and compared with the multivariable regression. The application of the GMDH network has been implemented in several fields [34][35][36][37][38][39][40][41][42][43][44]. The GMDH algorithm, as described in several references [34,35,46,53], utilizes multiple inputs to identify the best combination and generates a quadratic polynomial.…”

“…Among others, GMDH has been utilized for accurate log interval value estimation (Mohammed Ayoub (2014) [34]), permeability prediction by Alvin K. Mulashani (2019) [35] and Lidong Zhao (2023) [36], as well as permeability modeling and pore pressure analysis by Mathew Nkurlu (2020) [37]. Additionally, GMDH finds applications in cement compressive strength design (Edwin E. Nyakilla, 2023 [38]), rock deformation prediction (Li et al, 2020 [39]), bubble point pressure estimation by Fahd Saeed Alakbari (2022) and Mohammad Ayoub (2022) [40,41], gas viscosity determination, CO 2 emission modeling (Rezaei et al, 2020 and2018 [42,43]), the prediction of CO 2 adsorption by Zhou L. (2019) [44] and Li (2017) [45], forecasting stock indices, and modeling power and torque as demonstrated by Ahmadi (2015) [46] and Gao Guozhong (2023) [47], and the prediction of pore pressure by Mgimba (2023) [48].…”