Quantitative Method of Classification and Discrimination of a Porous Carbonate Reservoir Integrating K-means Clustering and Bayesian Theory

Reservoir classification is a key link in reservoir evaluation. However, traditional manual means are inefficient, subjective, and classification standards are not uniform. Therefore, taking the Mishrif Formation of the Western Iraq as an example, a new reservoir classification and discrimination method is established by using the K-means clustering method and the Bayesian discrimination method. These methods are applied to non-cored wells to calculate the discrimination accuracy of the reservoir type, and thus the main reasons for low accuracy of reservoir discrimination are clarified. The results show that the discrimination accuracy of reservoir type based on K-means clustering and Bayesian stepwise discrimination is strongly related to the accuracy of the core data. The discrimination accuracy rate of Type I, Type II, and Type V reservoirs is found to be significantly higher than that of Type III and Type IV reservoirs using the method of combining K-means clustering and Bayesian theory based on logging data. Although the recognition accuracy of the new methodology for the Type IV reservoir is low, with average accuracy the new method has reached more than 82% in the entire study area, which lays a good foundation for rapid and accurate discrimination of reservoir types and the fine evaluation of a reservoir.