Transitional Shale Reservoir Quality Evaluation based on Random Forest Algorithm: a case study of the Shanxi Formation, Eastern Ordos Basin

The Ordos Basin, characterized by its abundant transitional shale gas resources, plays a 11 significant role in Chinese oil and gas exploration industry. However, the complex sedimentary environment and lithofacies combination of transitional shale make it highly challenging for reservoir quality evaluation. Acknowledging the rapid development of artificial intelligence, particularly the extensive use of machine learning in geology, this study proposes a new approach to assess the quality of transitional shale reservoirs through the utilization of the Random Forest 16 algorithm (RF). Firstly, the lithology identification chart and reservoir quality evaluation standard were established using experimental data and logging curves, and the relevant datasets were constructed. Four logging curves (Acoustic curve (AC), Compensated Neutron curve (CNL), Density curve (DEN), Gamma Ray (GR)), which serve as input variables to reflect reservoir characteristics, were carefully selected, while reservoir quality classification was used as the output results. Subsequently, the RF model was constructed and trained using this dataset. By analyzing the confusion matrix, it was observed that the RF model achieved an impressive accuracy level of approximately 90%. The study confirmed RF's superiority through comparisons with five methods: factor analysis, Bayesian discriminant analysis, Gaussian Mixture Model (GMM), K-nearest neighbors (KNN), Gradient Boosting Decision Tree (GBDT). The comparison results revealed that the RF model exhibited high reliability and practical efficiency. Additionally, the RF model is utilized to predict the thickness of Type 1 reservoirs in the study area. The results demonstrated remarkable success in confirming production data, further emphasizing the proficiency of the RF within the field of machine learning for evaluating transitional shale reservoirs. This method presents a valuable tool for assessing transitional shale reservoirs.