Application Of Online Simulation In Power Plant Denitrification System

With the continuous development of simulator technology in thermal power plant,the role of the simulator is no longer limited to offline training,but can also be used to conduct research on the characteristics of field equipment,verification of new control schemes,etc.With the longterm operation of thermal power units,the characteristics of field equipments and devices will change due to wear,aging,coking,ash accumulation or catalyst activity reduction,etc.It is especially important to implement online simulation of the simulator,i.e.,the models in the simulator can be updated with the changes of field equipment characteristics.In this paper,the Selective Catalytic Reduction(SCR)flue gas denitrification system is studied,and the modeling method of the controlled object of the SCR denitrification automatic control system is studied based on the historical operation data of the SCR equipment: the model of the SCR flue gas denitrification system is established by using the Linear Parameter Varying(LPV)model and Back Propagation(BP)Neural Network,and a suitable prediction error threshold is designed to determine whether the model is mismatched and need to be update.The online simulation of the automatic control system on SCR equipment is basically realized.The main research contents of this paper are.(1)Has discussed the research background and significance of online simulation,introduced the current simulator and the method、research status of SCR flue gas denitrification system modeling,also analyzed the process flow and reaction principle of SCR flue gas denitrification system.(2)In view of the large amount of data and high dimensionality of variables in the historical operation data of power plants,this paper firstly pre-processes the operation data,and then uses Pearson correlation analysis method and mutual information method to determine the input variables and the delay time between input and output variables of the SCR flue gas denitrification system.(3)The SCR flue gas denitrification system is a complex nonlinear system,so the LPV model and Back Propagation Neural Network are used to model the SCR flue gas denitrification system,and the model mismatch evaluation index based on historical performance benchmark is designed as the condition and basis for judging whether the model is mismatched or not,and the model is automatically updated when the model is judged to be mismatched.The simulation results show that the prediction accuracy of the output variables of the model obtained by online simulation is greatly improved compared with that of the model obtained by offline simulation.(4)The model update in the online simulation is used to synchronize the model update in the controller of the advanced control system for denitrification,which improves the stability of the control loop and enhances the performance of the controller.