Logging interpretation method based on Bayesian Optimization XGBoost

With the increasing complexity of geological conditions and labor cost of oil exploitation, processing the obtained geological and logging interpretation information is vital before formal exploitation. The existing logging interpretation technology relies heavily on manual and expert experience, as well as traditional interpretation algorithm has inherent defects of easy over fitting and low accuracy, which is difficult to realize logging interpretation under complex geological conditions. To address these concerns, this paper presents a XGBoost model based on Bayesian optimization for logging interpretation, which used machine learning method to reduce manual dependence and can be used in complex geological conditions. The input features of the proposed model are selected through the correlation ranking of logging interpretation features, and the smaller correlation features are eliminated. The Bayesian optimization algorithm is introduced to optimize the model to the best state. To evaluate the performance of the proposed model, a series of experiments on logging data of sandy conglomerate reservoir are conducted. The results are compared with other machine learning algorithms and indicate that Bayesian optimization XGBoost (BO-XGBoost) is feasible and powerful in the field of logging interpretation.