نوع مقاله : مقاله پژوهشی
نویسندگان
دانشکده مهندسی معدن، دانشگاه صنعتی اصفهان، اصفهان، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Accurate prediction of pore pressure is a cornerstone of safe drilling operations, reservoir modeling, and hydrocarbon migration analysis. This study addresses the challenge of pre-drill pore pressure estimation in the carbonate Asmari Formation of the Kupal oil field, southwestern Iran, by integrating 3D seismic inversion with machine learning techniques. The methodology combines acoustic impedance models derived from post-stack seismic inversion with well-log data—including RFT pressure measurements, sonic logs, and density logs—to develop predictive models for pore pressure and identify zones of abnormal pressure. Three approaches were evaluated: multivariate regression, probabilistic neural networks, and multilayer feedforward neural networks.
The workflow began with seismic inversion using the Hampson Russell software to generate an acoustic impedance model, calibrated against well-derived impedance from sonic and density logs. Key seismic attributes, such as acoustic impedance, amplitude-weighted phase, and apparent polarity, were extracted and correlated with RFT-measured pore pressures from three wells. A fourth well served as a blind test to validate model accuracy. Comparative analysis demonstrated that the multilayer feedforward neural network approach yielded the most reliable predictions, achieving a validation correlation of 87% and a mean relative error of 2.8%, outperforming regression (85.9% correlation, 3.15% error) and probabilistic neural networks (77% correlation, 3.31% error).
Spatial analysis revealed systematic deviations from hydrostatic pressure trends, particularly in structurally compartmentalized zones along the Kupal anticline. Overpressure anomalies, likely associated with shale layers and restricted fluid communication, were identified at depths exceeding 2,020 milliseconds two-way time. These zones align with the field’s anticlinal geometry, suggesting a link between tectonic deformation, lithological heterogeneity, and fluid entrapment. The results underscore the value of 3D seismic data for resolving lateral pressure variability in heterogeneous carbonate reservoirs, where well-based methods are spatially limited.
The study highlights three key contributions: (1) a seismic-driven framework for pore pressure prediction that reduces reliance on well-dependent measurements, (2) empirical validation of neural networks for petrophysical parameter estimation in data-constrained settings, and (3) identification of overpressure mechanisms in the Asmari Formation, critical for mitigating drilling risks such as wellbore instability. By enabling spatially continuous pressure modeling, this approach enhances pre-drill planning and reservoir management in geologically complex basins. Future work could integrate additional seismic attributes and rock physics models to further refine predictive accuracy.
کلیدواژهها [English]
)