Latest revision as of 11:50, 11 February 2026

Abstract

The study of overpressure genesis mechanism is the foundation of hydrocarbon reservoir formation and pressure prediction research, and a thorough understanding of the formation and distribution patterns of hydrocarbon resources is essential for practical hydrocarbon exploration. However, the prediction of anomalous high-pressure genesis currently encounters numerous challenges, including the complexity of overpressure genesis, the superimposed effect of multiple mechanisms, and the dependence of traditional models on manual analysis, resulting in inefficiency and quantification challenges. To this end, this paper proposes a method for identifying and quantitatively evaluating stratigraphic overpressure mechanisms based on neural architecture search, to enable rapid and accurate quantitative evaluation of overpressure mechanisms. The results show that the main anomalous high-pressure genesis mechanisms in the target work area include undercompaction, fluid expansion and tectonic compression, with contribution rates of approximately 73% for undercompaction, and 9% and 18% for fluid expansion and tectonic compression, respectively. The model’s accuracy in the test set reaches 95.4%, significantly enhancing the identification accuracy of anomalous stratigraphic pressure genesis and its superposition relationships. The innovation of this paper lies in the combination of wave velocity-density rendezvous map with clustering algorithm and neural architecture search algorithm, offering an efficient approach to identify multiple overpressure genesis mechanisms and predict pore pressure through machine learning algorithms, which is of great theoretical significance and practical application value.OPEN ACCESS Received: 08/08/2025 Accepted: 14/10/2025 Published: 03/02/2026

Document