Prediction Of NO_x Concentration Based On Large Delay Characteristics Of SCR Denitration System And Its Application

As an important part of my country’s energy structure,thermal power generation will still occupy an important position in my country’s power production structure in the next few decades.With the improvement of people’s awareness of environmental protection,thermal power plants have begun to implement ultra-low emission standards for pollutants such as NO_x generated in the process of thermal power production.Under the requirements of current standards,it is very important whether the SCR flue gas denitrification device of thermal power plants can achieve precise ammonia injection.Insufficient ammonia injection will cause the NO_x content in the flue gas to exceed the standard,and excessive ammonia injection will affect the overall operation safety of the unit.In this paper,by establishing a prediction model that can accurately predict the NO_x concentration of the SCR inlet flue gas under all operating conditions,as a feedforward signal to guide the SCR ammonia injection,the precise ammonia injection control of the SCR is realized.Firstly,aiming at the large delay characteristic of SCR caused by the delay of CEMS measurement,an orthogonal thermal state test was designed and carried out to measure the delay time of CEMS measurement at the SCR inlet to provide data support for the NO_x prediction model considering the time effect.At the same time,the working conditions during the design test can basically cover the working condition range of the boiler during normal operation,and provide comprehensive test data support for the establishment of a NO_x generation prediction model that can operate under all working conditions.The XGBoost algorithm was selected,based on the boiler operation data during the orthogonal thermal state test,adding time effect parameters,and taking into account the delay time measured by CEMS,a model that could accurately predict the NO_x content of the SCR inlet flue gas under all operating conditions was established.At the same time,the influence of different hyperparameter changes in the XGBoost algorithm on the accuracy of the model is studied,and the genetic algorithm is used to optimize the hyperparameters of the XGBoost model,the final NO_x concentration prediction model can shorten the measurement delay of CEMS by nearly 80%.The established NO_x concentration prediction model is set on the server that can interact with the real-time data of DCS,and the real-time prediction of the NO_x concentration of the target unit in actual operation is carried out.On the basis of proving that the predicted results are accurate and reliable,the predicted value of NO_x concentration is added to the ammonia injection feedforward control,and finally the precise ammonia injection of SCR is realized.During the official operation of the power plant,the NO_x concentration burr in the exhaust flue gas was significantly reduced,the fluctuation of the SCR ammonia injection amount was obviously smooth,and the blockage of the air preheater caused by ammonia escape was significantly improved.