Font Size: a A A

Modeling And Optimization Strategy Of Ammonia Injection Of SCR Denitration System

Posted on:2021-12-19Degree:MasterType:Thesis
Country:ChinaCandidate:D B RaoFull Text:PDF
GTID:2491306104984429Subject:Thermal Engineering
Abstract/Summary:
In order to stabilize the impact of new energy on the power grid,coal-fired power stations have undertaken more tasks of changing load.Wide load flexible operation brings a series of economic and safe operation problems to the power station.SCR denitration system has significant characteristics of large inertia,long delay and strong nonlinearity.In different load sections,the inertia of the system is quite different.Therefore,traditional linear PID control has been difficult to meet the requirements of precise ammonia injection under wide load and flexible operation and is unable to balance the safety,economy and environmental protection of the SCR system.After studying the mechanism modeling method,the concept of surface-internal adsorbed NH3 was proposed in this paper,and the mechanism model suitable for the actual data of power plant was established.To solve the problem of model failure during long-term operation,selective ensemble model library algorithm was proposed.A kind of model predictive control strategy based on the model library was established to reduce ammonia injection and escape under the condition that outlet NOx doesn’t exceed the standard.According to the mechanism model,this paper proposed the concept of surface-internal adsorbed NH3 and improved the reaction dynamic equations.And based on the equations,the dynamic mechanism model of SCR denitrification system was established by using one day operation data of SCR denitration system of a 660MW unit.The study found that the improved model has a better fitting effect,and the mean square error on the test set had increased from16.68%to 9.90%.In order to keep the data model effective in the long-term running process,a method named selective ensemble model library was proposed according to the similarity of data distribution in a short time.The model library included construction,combination and update strategies.After training and testing with 50 days of operational data,it was found that the selective ensemble model library’s MRE can still maintain 3.32%when the traditional sliding-window method’s had a sudden drop to 8.20%.A new ammonia injection control system was composed of model predictive control and selective ensemble model library.It was established to verify the effect in ammonia injection control.The optimal ammonia injection quantity calculated by the strategy was input into the SCR simulation reactor established by the mechanism model.The study showed that under stable load,the value of the new system was more stable,its average value was 40.97mg/m3,which was closer to the set value than 42.35mg/m3 of PID,and its variance was 6.68,which was far less than 12.82 of PID.Under variable load the value was controlled well.The maximum value is 52.02mg/m3,which was lower than 67.08mg/m3 of PID,and the number of exceedances was 15 which was far lower than 139 of PID.
Keywords/Search Tags:coal fired power station, selective catalytic reduction, mechanism modeling, machine learning, model predictive control
Related items