An ACO-LSSVM-Based Identification Method of Fracture-Vug Fillings

Abstract

The types of fracture-vug fillings in carbonate reservoirs play a key role in the storage and transportation of hydrocarbons. Accurate identification of fracture-vug fillings is crucial for evaluating reservoir quality. For the deep carbonate reservoirs of Cambrian Longwangmiao Formation in Muxi Block, we summarized the response characteristics of fillings from coring data, conventional logging data, and electric imaging logging data, and proposed an ant colony algorithm (ACO)-based optimization method of the important parameters of the least-squares support vector machine (LSSVM) for the fracture-vug fillings identification (ACO-LSSVM). Then, based on the refined core descriptions, ACO-LSSVM was applied and validated. The validation results showed that the fracture-vug fillings could be mainly classified as mud fillings, quartz fillings, and carbonate mineral fillings. In addition, reconstructed natural gamma relative value, sonic ratio, density ratio, and other combination curves gave the more obvious responses to different types of fracture-vug fillings than conventional logging curves. Compared to traditional support vector machines (SVM) and random forest (RF) methods, the ACO-LSSVM identification method demonstrates superior performance, with an overall accuracy improvement of more than 10% over the random forest method. Additionally, it exhibits better robustness and a positive feedback mechanism, allowing it to find the optimal solution in a shorter time. The accurate identification of fracture-vug fillings provided the foundation for valid fracture-vug parameter extraction and accurate evaluation of carbonate reservoirs.