Fluid Property Identification of CatBoost Model Based on Grey Wolf Optimization Algorithm

Low porosity and permeability reservoirs play an important role in the development of oil and gas resources. This kind of reservoir has strong heterogeneity, complex pore structure, and unclear characteristics of log responses. Conventional methods, therefore, are difficult to identify its fluid properties. This study explores the adoption of GWO-CatBoost model for fluid properties’ identification. Firstly, the GWO algorithm is used to globally optimize the core parameters of the CatBoost model, thereby enhancing the generalization performance and predictive accuracy of the model. Secondly, the logging curves of the study area were input into the optimized model to achieve high-precision fluid property identification. To test the validity of this method, this study selected CatBoost model, random forest (RF), logistic regression (LR), and LightGBM model for comparative experiments. The findings show that the GWO-CatBoost model has superior accuracy of fluid identification, reaching 92.78%, which is significantly improved compared with other models. This provides a new idea for fluid identification in low porosity and permeability reservoirs.