Optuna-XGBoost porosity prediction for logging profiles of carbonate formations

Carbonate reservoirs occupy an important position in the world’s oil and gas reserves. At the same time, carbonate reservoirs are also one of the most important reservoir types in China, which play an important role in China’s oil and gas energy development strategy, and it is of great significance to effectively utilize the existing logging curves to maximize the prediction of relevant reservoir information. Porosity is an important parameter for reservoir evaluation. The prediction of porosity based on logging data is simple and fast compared to the method of core sample detection. However, the machine learning method has too many hyperparameters and the model training is difficult to optimize. To improve the accuracy of model prediction, a porosity prediction method based on Optuna-XGBoost is developed in this study. The optimization models of PSO-XGBoost and Optuna-XGBoost models of XGBoost were established, and it was found that Optuna-XGBoost based on the Optuna optimization framework has better prediction performance in comparison; finally, the optimization models of SVR, GBDT, and XGBoost were established using the Optuna hyperparameter optimization framework. The experimental results show that the Optuna-XGBoost model after Optuna optimization has better prediction performance compared with Optuna-SVR and OptunaGBDT, and can well overcome the problems of time-consuming and low prediction accuracy of traditional parameter optimization methods. Therefore, the prediction method based on OptunaXGBoost is an effective method for predicting porosity, which has certain applicability and reference value in log evaluation research.