In order to study the dynamic characteristics of supercritical once-through boiler units and design a coordinated control algorithm for it

a supercritical 600 MW coal-fired once-through boiler unit is taken as the research object and a mechanism model of the unit is established. In the model

the complex heat absorption equation and steam pressure difference equation in boiler section

as well as the superheated steam flow equation and unit load coefficient equation in turbine section are fitted by the long short-term memory(LSTM) neural network with real operational data from the unit given the algorithm's superiority in handling sequence data with long-term dependencies. This established four sub-models of the equations that can capture the operating characteristics of the unit. For the dynamic parameters in the mechanism model

an improved differential evolution algorithm is proposed to identify. After this

a complete state equation model is obtained. The established model is verified through open-loop step disturbance testing and closed-loop simulation upon historical operation data. The results show that the established model can accurately reflect the dynamic operating characteristics of the unit. The mean absolute percent error of the main steam pressure

steam enthalpy in separator and unit load are all less than 1.76%

which means the established model has high accuracy and can be used for research on coordinated control algorithms of supercritical units.