Identification And Control Of Coal-Fired Power Plant SCR De-NOx System Based On Historical Operation Data

The selective catalytic reduction(SCR)system of coal-fired power plant is a time-varying nonlinear system with great inertia and large time delay.The present control system of SCR is difficult to achieve a satisfying performance using a combination of feed-back and feed-forward control strategies based on a fixed mixing molar ratio of NH3 to NOx.Two improved control strategies have been proposed to handle the problem,namely Smith predictor control and model predictive control.However the performance of the improved control strategies would be poor when the identification model mismatches.Currently,SCR system model in coal-fired power plant is identified by industrial field experiments.These experiments are costly and undesirably disturb SCR system.Considering the aforementioned issues,an identification method of SCR system is developed based on historical operation data.Additionally,the design of SCR control system is studied by using the identification models.Finally,the effect of the model mismatch on control performance is discussed.The NOx formation mechanisms,existing De-NOx technologies and,the working principles of SCR system are introduced.According to the SCR system characteristics in coal-fired power plant,the problems of the present control scheme are analyzed.Two improved control strategies are introduced to overcome the existing problems in SCR control system.In addition,the effect of model mismatch on the performance of the improved control approaches is discussed.The defects of classical identification methods are demonstrated from the view of practical industrial applications.Furthermore,the asymptotic method is introduced,which can reduce the effect of the industrial disturbances on the accuracy of identification model.Based on mining techniques,the condition number of Fisher information matrix has been proposed to remove the outlier data and the redundant data which do not have sufficient excitation to determine a model.The asymptotic identification method is applied to identify the models of SCR system with the filtered historical data.It is shown that the proposed method is accurate and robust with a simple structure and a relatively low amount of calculation.According to the identification models of SCR system,the PID controllers are tuned by the Cohen-Coon method.Through simulation experiments,the control system of SCR is designed to deal with the existing problems in practical control system.It is shown that the cascade control scheme is efficient to eliminate disturbances in ammonia spraying system.Nevetheless,it does not work on the process with a large time delay.To improve the control performance,Smith predictor control and model predictive control are adopted to settle the problem of large time delay.Finally,the effects of the model mismatch on the performance of improved control schemes are studied.The simulation results showed that an accurate identification model is helpful in the design of control system,improving the stability and economy of SCR control system.