Machine learning based on a swarm intelligence algorithm and explainable AI for the prediction of reservoir temperature

Reservoir temperature is a critical parameter for geothermal fluids, and machine learning techniques can leverage geological and geothermal exploration data to identify patterns in subsurface geothermal reservoir temperatures. In this study, we propose a hybrid model that integrates five machine learning models with the Swarm Intelligence algorithm (Ant-Lion Optimization, ALO) to predict geothermal reservoir temperatures. The model is applied to geothermal data from the central depression area of the southern Songliao Basin, China. Our results show that the ALO-GPR model achieves the highest prediction accuracy, with an R2 of 0.99 and an optimal RMSE of 0.038. To assess the impact of input variables on model performance, we employed SHAPley Additive ExPlanations (SHAP), which highlighted the significant influence of reservoir depth (D), sodium and potassium (Na + K), and boron (B) on temperature predictions. These findings demonstrate the potential of meta-heuristic algorithms to enhance machine learning models' predictive capabilities for geothermal reservoir temperatures. Additionally, combining machine learning models with interpretable algorithms provides valuable insights for developing effective geothermal resource management strategies.