No abstract

Summary With the affirmative advancements in clean coal technologies toward more environmentally friendly power systems, it has become critical to gain a comprehensive understanding of the modeling philosophies of such technologies. Supercritical (SC) and ultra‐supercritical (USC) technologies are among the leading options of clean coal technologies, and it is widely acknowledged that they provide flexible power generation for grid electrical demand while sustaining cleaner operations. This paper presents detailed qualitative and quantitative comparisons of the most salient modeling methodologies of a 600 MW SC generation unit which study the once‐through operation to provide future researchers with a better understanding of this field and be used as a reference for more advanced strategies and decision‐making problems, such as efficiency improvements, emissions reduction, and control design for industry personnel, by determining the key characteristics and drawbacks of each modeling technique. These methods are first‐principles or physics‐based modeling, state‐space system identification (SI) modeling using prediction error minimization, and artificial neural networks (ANN) modeling using the non‐linear autoregressive exogenous input (NARX). The comparison was found to be effective in recognizing the diversity in the computational capabilities of the three developed models. The simulation errors for the adopted output responses are presented to quantify the comparison and justify the superiority and imperfection of each method in terms of accuracy. Other qualitative factors for comparison include simplicity and physics‐based sagacity. It has been eventually deduced that in terms of accuracy, the ANN‐based model using NARX is the most accurate model and has the best performance for black‐box modeling, whereas the linearized state‐space model using SI modeling is the most suitable one for designing linear controllers, and the physical model is superior for physical interpretation and stability studies. The control applications of each technique have been highlighted as well. Highlights Three different models are presented for a 600 MW cleaner SC boiler‐turbine‐generator unit that covers the once‐through mode. A comprehensive realization of the three modeling techniques is presented. Simulation results have shown the suitability of the three methods for this specific application with different capabilities, computational burdens, and possible control methods. A comparative study is then reported and discussed in detail, which can be used as a future educational guide for future researchers.