Forecast of NOx Emissions for a 660MW Coal-Fired Boiler with Multilayered Gradient Boosting Decision Tree Considering...

Area of Science:

• Environmental Science
• Chemical Engineering
• Data Science

Background:

• Nitrogen oxides (NOx) are key pollutants from coal-fired power plants.
• Accurate NOx prediction is vital for optimizing boiler control and reducing emissions.

Purpose of the Study:

• To develop a data-driven methodology for predicting NOx concentrations at the boiler outlet.
• To enhance the accuracy of NOx emissions prediction in coal-fired power plants.

Main Methods:

• Utilized Kernel Independent Component Analysis (KICA) for nonlinear correlation elimination.
• Integrated physically-grounded variables with KICA-extracted features.
• Employed Robust Gaussian Mixture Model (RGMM) for identifying operational modes.
• Developed mode-specific multilayered Gradient Boosting Decision Tree (mGBDT) models.
• Optimized hyperparameters using Particle Swarm Optimization (PSO) and 10-fold cross-validation.

Main Results:

• Achieved a coefficient of determination (R²) of 0.947.
• Obtained a root-mean-square error (RMSE) of 6.09 mg/m³.
• Reported a mean absolute error (MAE) of 4.009 mg/m³.
• Demonstrated superior performance compared to five other models.

Conclusions:

• The proposed mGBDT framework effectively predicts NOx emissions.
• The methodology enhances boiler control and emission reduction strategies.
• This data-driven approach offers a significant improvement in NOx prediction accuracy.