Model Predictive Control Disturbance Rejection And Model Repair Strategies For SCR Denitrification System

In order to effectively improve the atmospheric environment,relevant government departments in China have formulated strict emission standards for pollutants from large industrial processes,and denitrification has become an important part of flue gas treatment for coal-fired power plants.As a typical industrial process with large inertia and large delay,the selective catalytic reduction(SCR)denitrification system often can not achieve satisfactory control performance under the traditional PID based controllers.For its full-developed theories and excellent nature,model predictive control(MPC)has been applied in SCR denitrification system.With the proposal of carbon peak,carbon neutralization and other goals,thermal power units will undertake the task of peak load regulation and frequency regulation increasingly,and their operation conditions are becoming more and more complex,which brings new challenges to the control of SCR denitration system.How to improve the performance of MPC has become the key to ensure the stable and economic operation of SCR denitration system.Two approaches are taken to increase the performance of MPC,which are improving its disturbance rejection ability and ensuring the model accuracy.In the aspect of disturbance rejection of MPC,an augmented state space model is proposed for disturbance estimation.An adaptive black box model is then established for the estimated disturbance sequence utilizing online extreme learning machine with kernels,which makes full use of the trend information to effectively reject the disturbance with strong regularity and relatively obvious trend.In terms of model online repair,a model-plant mismatch monitoring method based on slow feature analysis is proposed.It can effectively monitor the model-plant mismatch by analyzing the variation trends of the deviation between model and plant.Simulation research shows its effectiveness compared with several typical algorithms.Combined with the closed-loop subspace identification algorithm,the proposed scheme can effectively repair the model and improve the control performance.For the industrial application of the proposed advanced control schemes in SCR denitrification system,a modeling scheme of SCR process based on converted ammonia injection quantity is proposed,which can transform the SCR into quasilinear process,bringing convenience to modeling and control.The processing methods of nonlinearity of valves are analyzed and discussed.The control software is developed based on Python,and the corresponding modification scheme of configuration diagram in DCS is designed.