Research On Compound Modeling Of Supercritical Unit Steam Temperature System

The modeling,simulation and dynamic characteristics of the controlled object are the basis of the design of the process control system.Only by clearly understanding the dynamic characteristics of the controlled object,it is possible to design a control system with excellent performance.After more than ten years of development,the supercritical unit has become the main unit of the power grid.Compared with the steam drum furnace unit,its steam temperature system is more complicated and more difficult to control.How to establish a more accurate and practical supercritical unit steam temperature system model has become one of the research hotspots in the field of thermal engineering in recent years.In this paper,the 1000 MW supercritical unit of a power plant of Guodian is taken as the research object,and the composite modeling method of steam temperature system is studied.It is intended to establish a composite model with clear physical meaning and more accurate reflection of the dynamic characteristics of the steam temperature system.The steam temperature system is divided into three typical objects: superheater,reheater and water spray desuperheater,and their dynamic models are established respectively.Firstly,the mechanism model of the research object is established,and the structure of the model,the coefficient to be identified and the process variables affecting the output parameters of the model are determined.Then,the neural network model for determining the coefficient of the mechanism model is established,and the random approximation of the simultaneous disturbance is used by using the field data.The SPSA algorithm determines the weight of the neural network.Finally,the mechanism model of the object and the neural network model for the computer model coefficient form a composite model of the steam temperature system,which is established by field data pairs under different time periods and different working conditions.The model is validated by simulation.The research results show that the composite model of steam temperature system can achieve higher precision.Under various operating conditions,its dynamic characteristics are close to the actual unit,and the physical meaning of the model is clear and the scope of application is large.The SPSA algorithm can make the composite model training process converge well,especially for the composite model using the mechanism model as the main model and the neural network model as the coefficient model.