Strategic-level performance enhancement of a 660 MWe supercritical power plant and emissions reduction by AI approach

•The power plant heat rate is modeled by system-level and environmental factors.•Parametric significance order on the power plant heat rate is determined.•AI model-based knowledge extraction from the operational data is presented.•Significant improvement in thermal efficiency and fuel savings are obtained.•Annual reduction in CO2, SO2, CH4, N2O, and Hg emissions are estimated. Power plant heat rate is a plant level performance parameter that indicates the economy of power production, equipment’s safety, and availability. In this paper, seven operating parameters, including the performance indices of integrated energy devices and the environmental conditions are incorporated for modeling the power plant heat rate by Artificial Neural Network (ANN), Support Vector Machine (SVM), and automated machine learning (AutoML) approach. The parametric significance order is determined by ANN and SVM-based Monte Carlo analytics and other machine learning-driven algorithms. Subsequently, the best-performing model is selected based on the external validation test and deployed for knowledge mining purposes. The improvement in the power plant heat rate by the parametric adjustment is achieved and subsequently, up to 3.12 percentage point (pp) increase in the thermal efficiency of the power plant is confirmed. Moreover, the fuel savings corresponding to the improved power plant heat rate are also calculated at three power generation modes. Their equivalence to an annual reduction in emissions is quantified. It is estimated that the accumulated reduction in CO2, SO2, CH4, N2O, and Hg emissions, i.e., 288.2 kilo tons / year (kt/y), can be achieved under 3.15% improvement in the power plant heat rate, corresponding to 75% power generation mode.