Soft Sensing And Predictive Control Of Denitration Reaction Process

It is significant for China to acquire the experience which Western developed countries have learned in the process of industrialization and develop in the direction of the new industrialization with green and sustainable development.Environmental protection departments have strengthened restrictions on the generation and discharge of pollutants from industrial processes and have formulated special standards for the discharge of pollutants from coal-fired power plants.Selective catalytic reduction(SCR)refers to the reaction of reducing agent and nitrogen oxides(NO_x)in flue gas under the action of a catalyst to produce pollution-free nitrogen and water.Selective Catalytic Reduction flue gas denitration technology is widely used in the research of nitrogen oxide emission reduction in thermal power plants due to its extreme purification rate,reliable operation,and no secondary pollution.The amount of ammonia injection is the most important control variable.Too little ammonia injection will result in an insufficient reaction.Low denitrification efficiency and too much ammonia injection result in increasing the amount of ammonia escape and increasing the cost of denitration.In this paper,the SCR denitrification system of domestic large-scale coal-fired power plants is considered as the research object.Using the historical data in the Distributed Control System to analyze the mechanism of the denitrification process,the author establishes an accurate export nitrogen oxide concentration model and proposes a new control strategy on this basis to achieve predictive control of the denitration process which takes the comprehensive cost into account.The main contents include:(1)According to the structure and reaction principle of denitration system in thermal power plant,the mechanism model of SCR denitration system is constructed by analyzing the denitration reaction mechanism.Considering that the operating characteristics of the unit under changing conditions,this paper selects inlet temperature,inlet pressure,and unit load,and uses a sliding window strategy to detect steady-state conditions.Combined with outlier elimination,the sample data is selected.Finally,the model parameters are indicated with a particle swarm optimization algorithm.(2)The soft sensor model of denitration system outlet concentration is established.For the variable selection problem in the process of soft sensor technology implementation,the correlation between an auxiliary variable and dominant variable outlet concentration,the redundancy between auxiliary variable and auxiliary variable are considered,and the idea of k-nearest neighbor mutual information is introduced to realize the selection of candidate variables.Adaptive weighted least squares support vector machine(AWLS-SVM)is used to train the samples,and the soft sensing model of outlet concentration is established.(3)Aiming at the problems of measurement delay and inaccurate PID control of traditional denitration system in thermal power plant,the denitration system cannot be controlled timely and effectively.Based on the soft sensing model and mechanism prediction model of outlet concentration,this paper designs a multi-objective predictive control strategy.The design is aimed to reduce the outlet concentration of the denitration system and the cost of real-time denitration.Compared with the traditional control method,it shows the effectiveness of the strategy.