Optimization of Heat Rate and Greenhouse Gas Emission Reduction at Coal-Fired Power Plants in Indonesia Through Machine Learning Modeling

Abstract

This study aims to develop predictive models for the heat rate of coal-fired steam power plants (CFSPPs) in Indonesia using various machine learning techniques and to identify factors influencing greenhouse gas emissions, specifically CO₂. Techniques used include Linear Regression, Lasso Regression, Polynomial Regression, Ridge Regression, Support Vector Regression, Random Forest Regression, Gradient Boosting Regression, Elastic Net Regression, AdaBoost Regression, Neural Network Regression, Decision Tree Regression, and Extra Trees Regression. The data consists of 468 performance test results from CFSPPs, covering operational parameters such as boiler type, ambient temperature, flue gas temperature, and unburned carbon. Analysis shows that the Extra Trees Regression model provides the best performance with an R-squared value of 0.947, MAE of 133.648, MSE of 34694.478, and RMSE of 186.265 for heat rate modeling, and an R-squared value of 0.993, MAE of 21.02, MSE of 1402.858, and RMSE of 37.455 for CO₂ emissions modeling, demonstrating high accuracy and good generalization. Significant factors influencing the heat rate include Gross Power Output (GPO), Net Power Output (NPO), load percentage, boiler type, coal HHV, coal consumption, and operational duration. This model is implemented using the Postman application for real-time heat rate and CO₂ emissions prediction, facilitating integration with CFSPP’s operational systems. The research results indicate that the application of machine learning can improve energy efficiency and reduce CO₂ emissions, supporting Indonesia's Nationally Determined Contribution (NDC) targets. This study provides new insights into the application of machine learning in the power generation industry and offers recommendations for further implementation and research.