Font Size: a A A

Low NO_x Emission Optimization For Power Plants Based On Neural Network And Predictive Control

Posted on:2019-01-07Degree:MasterType:Thesis
Country:ChinaCandidate:Z LiFull Text:PDF
GTID:2371330548489286Subject:Detection Technology and Automation
Abstract/Summary:PDF Full Text Request
During the period of 13 th Five-Year,the state has raised higher requirements for the control of air pollution.The flue gas of the power plant is one of the main factors of air pollution,so the optimization of the denitrification system of the power plant has been paid more attention.At present,the control strategy of the power plant denitrification system mostly adopts the compound control method,this is a control strategy which uses the mole ratio as feedforward with the feedback of PID calculation.However,due to the complexity of the field environment,there are many problems in the measurement of NO_x concentration of reactor inlet.Because of the large delay,inertia and nonlinearity of controlled object,the control effect of PID control mode is not good.In view of the measurement of the concentration of NO_x in the entrance of SCR reactor,a soft measurement model of NO_x concentration based on BP is proposed.Based on the historical data of the power plant,a soft measurement model is established.In this paper,a model aided variable filtering algorithm based on mutual information is improved and applied to the selection of auxiliary variables.In order to reduce the training time of the BP network,the genetic algorithm is used to optimize BP network.The sensor model solves the measurement problems caused by the purge of reactor inlet during the operation of the unit and the problem slow response of NO_x concentration measurement in the gas.And this model can calculate the NO_x concentration immediately according to the working condition parameters.In order to solve the problem of bad effect of PID control,dynamic matrix predictive control is used.First,the recursive least square algorithm is used to identify the SCR reactor model of a unit based on the field operation data.On the basis of this model,the control effect of PID control and dynamic matrix control is compared.Dynamic matrix control effectively solves the problem of large delay and large inertia.Because of the rolling optimization function of predictive control,it can also overcome the nonlinear problem of the system to a certain extent.At the end,a complex control strategy which combines the mole ratio feedforward calculation and the dynamic matrix feedback is introduced to a practical example of a power plant.It reduced the crystallization degree of the air preheater of the unit,avoided the frequent transient over standard of the flue gas emitted from the unit.The practice has proved that this control strategy has a good effect in the SCR denitration system.
Keywords/Search Tags:SCR denitrification system, Neural Network, Predictive control, Denitrification optimization
PDF Full Text Request
Related items