Exploration
Poorandokht Soltani; Amin Roshandel Kahoo; Hamid Hassanpour
Abstract
Seismic methods are among the primary and most effective techniques for hydrocarbon exploration, as they enable comprehensive imaging and interpretation of the Earth's subsurface. However, accurate interpretation of seismic data requires detailed analysis of geological structures, often involving complex ...
Read More
Seismic methods are among the primary and most effective techniques for hydrocarbon exploration, as they enable comprehensive imaging and interpretation of the Earth's subsurface. However, accurate interpretation of seismic data requires detailed analysis of geological structures, often involving complex and subjective decision-making processes. Constructing an initial geological model that aligns with seismic observations is a critical first step, but it is inherently non-unique and heavily influenced by the interpreter’s experience and preferences. Among various subsurface structures, salt domes are of particular interest due to their unique physical characteristics and their critical role in hydrocarbon entrapment, drilling risk management, and subsurface storage applications. Their distinct seismic textures, compared to surrounding sediments, make them identifiable using seismic texture attributes. Nevertheless, the manual delineation of salt dome geobody is a time-consuming and potentially error-prone task, especially given the volume, redundancy, and complexity of the seismic attributes used. To overcome these challenges, we propose a novel unsupervised framework for automatically identifying salt dome geobody in 2D seismic sections. The method begins by extracting a diverse set of seismic texture attributes, including both conventional attributes and novel texture descriptors derived from advanced image analysis techniques. Following attribute extraction, a attribute selection phase using techniques such as Laplacian Score is employed to eliminate redundant, irrelevant, or highly correlated attributes, thereby enhancing model efficiency and interpretability. The reduced set of relevant attributes is then used as input for clustering algorithms based on metaheuristic optimization techniques. These algorithms aim to partition the seismic data into meaningful clusters that correspond to geological attributes, particularly salt domes. Validation against multiple expert interpretations demonstrates the robustness and high accuracy of the proposed method. Results emphasize the capability of unsupervised clustering approaches especially those guided by metaheuristic strategies—in reducing interpretation uncertainty and improving segmentation quality.
Andisheh Alimoradi; Ali Moradzadeh; Mohammad Reza Bakhtiari
Abstract
This paper proposes a method for the prediction of pore size values in hydrocarbon reservoirs using 3D seismic data. To this end, an actual carbonate oil field in the south-western part ofIranwas selected. Taking real geological conditions into account, different models of reservoir were constructed ...
Read More
This paper proposes a method for the prediction of pore size values in hydrocarbon reservoirs using 3D seismic data. To this end, an actual carbonate oil field in the south-western part ofIranwas selected. Taking real geological conditions into account, different models of reservoir were constructed for a range of viable pore size values. Seismic surveying was performed next on these models. From seismic response of the models, a large number of seismic attributes were identified as candidates for pore size estimation. Classes of attributes such as energy, instantaneous, and frequency attributes were included amongst others. Applying sensitivity analysis, we determined Instantaneous Amplitude and asymmetry as the two most significant attributes. These were subsequently used in our machine learning algorithms. In particular, we used feed-forward artificial neural networks (FNN) and support vector regression machines (SVR) to develop relationships between the known attributes and pore size values in a given setting. The FNN consists of twenty one neurons in a single hidden layer and the SVR method uses a Gaussian radial basis function. Compared with real values from the well data, we observed that SVM performs better than FNN due to its better handling of noise and model complexity.
