Enhancing the Performance of Gas Turbines Predictive Emissions Monitoring System: A Stacked Ensemble Machine Learning Models

Climate change, driven by industrial greenhouse gas emissions, poses significant challenges to public health, the environment, and the economy. Predictive Emissions Monitoring Systems (PEMS) have evolved through advancements in machine learning, enabling improved accuracy and reliability in real-time emissions predictions. However, these systems face key limitations, including operational challenges, economic constraints, data quality issues, and the inherent weaknesses of individual models in achieving high accuracy and generalization. This study proposes a Stacked Ensemble Machine (SEM) approach, combining base learners (Extra Trees Regressor, Decision Tree, Random Forest) with a meta-learner (Linear Regression) to overcome these challenges. The base learners achieved superior predictive accuracy with minimal error metrics (RMSE, MSE, MAE, MAPE: 0.002–0.009) and high score of R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, while the meta-learner aggregated predictions effectively. The stacked ensemble method demonstrated enhanced generalization, error minimization, and robust predictive performance, offering a balanced and innovative solution for emissions monitoring. Moreover, the study highlights the adaptability of the SEM framework to various industrial applications, including power generation, oil and gas operations, and other emissions-intensive sectors. By ensuring precise monitoring and compliance with environmental regulations, the SEM approach offers a cost-effective and scalable solution for managing industrial emissions. In conclusion, the proposed SEM framework demonstrates significant potential in advancing the field of emissions monitoring. By combining the strengths of multiple machine learning models, it offers a balanced, innovative, and reliable approach to tackling the complex challenges of emissions prediction, thereby contributing to global efforts in mitigating climate change. Future research could explore the SEM model’s applicability to other emission sources or different industrial contexts to further validate its robustness and scalability.