MULTI-OBJECTIVE OPTIMIZATION OF BOILER COMBUSTION EFFICIENCY AND EMISSIONS USING GENETIC ALGORITHM AND RECURRENT NEURAL NETWORK IN 660-MW COAL-FIRED POWER PLANT

Indonesia has demonstrated a firm commitment to achieving Net Zero Emissions (NZE) by 2060. The implementation of diverse strategies, such as the application of biomass co-firing technology in coal-based steam power plants, demonstrates this commitment. This study focuses on the 660 MW supercritical coal-fired boiler as the object of investigation. The key problem addressed in this research is the unstable combustion performance due to the dynamic and nonlinear interactions among operational variables under biomass co-firing conditions. These fluctuations can negatively impact boiler efficiency, CO₂ emissions, and the plant’s capability factor. The study proposes a dynamic multi-objective optimization framework using a Recurrent Neural Network (RNN), Response Surface Methodology (RSM), and a Multi-Objective Genetic Algorithm (MOGA) to enhance performance reliability and support Indonesia’s transition to cleaner energy sources. The findings indicate that the RNN model exhibited superior prediction accuracy compared to the RSM, with a Root Mean Square Error (RMSE) value of 0.1523% for boiler efficiency, 1.6993% for CO₂ emissions, and 0.5284% for the capability factor. The MOGA optimization exhibited an enhancement in boiler efficiency from 86.6793% to 87.32%, a reduction in CO₂ emissions from 114.213 mg/Nm³ to 53.972 mg/Nm³, and an augmentation in the capability factor from 87.9% to 89.32%. Furthermore, coal consumption is reduced to 51,524 tons per hour, which can generate operational cost savings of IDR 1.34 billion per day. The RNN and MOGA-based approaches have been demonstrated to be more effective than RSM for optimizing boiler combustion. This method is important for developing a strategy to improve the efficiency of the combustion process in boilers in coal-fired power plants.