Model Research For SCR System Monitoring And Management Platform Based On Internet Of Things

Currently application of environment on-line monitoring and analysis system has been taken more and more seriously in China.For instance, flue gas desulfurization monitoring platform has been put into engineering application in the Inner Mongolia Autonomous Region which can predict the real operating data accurately and identify abnormal situation of desulfurization equipments.Based on application experience of desulfurization platform and massive data source of local environment network,NOx reduction models had been built which can be used to monitor the industrial equipments,calculate total NOx emission,calculate reduction of emission and charge for pollutants emission.The model can replace traditional way of management which is simple and ineffective.As a result,management and application ability of atmospheric stationary pollution source emission monitoring platform in Inner Mongolia would be enhanced.Qualitative analysis model was built based on operating data, which can filter important NOx reduction operating data according to the rule from loose to strict.There were three parts contained in qualitative analysis model.First part,the range of important operating parameters by different load was concluded according to performance data of NOx reduction equipment collected from different plants. Second part, by using Pearson correlation method our research calculated the correlation of operating parameters and proposed confidence interval of correlation.Third part, analysed logical relationship among operating parameters according to the dynamic process of the NOx reduction operation as a supplement to the validity test of operating data.Quatitative analysis model which chose JinQiao power plant as basis was built based on NOx production and emission. NOx production models included SCR inlet NOx concentration model and NOx production model. SCR inlet NOx concentration model was built based on neural networks.Main factors that contribute to NOx production was confirmed at first as input parameters of the model.Subsequently, model parameters such as transfer function,train function and number of nodes in hidden layer was confirmed.The model was trained based on the operating data of 1# unit of JinQiao power plant and was used to predict inlet NOx concentration of 1#,2#units of JinQiao power plant after trained.The result revealed that the model can predict operating data of 1# unit accurately compared with worse prediction performance of 2# unit,which demonstrated it was necessary to use corresponding unit operating conditions to build accurate model based on the same model structure. NOx production was the product of inlet concentration and flue gas flux.NOx emission models included efficiency of NOx reduction model,NOx reduction model,NOx emission model and SCR outlet NOx concentration model. Mass transfer of reaction components in the channel and micropore of catalyst was described by Fick's law. Partial differential equations related with concentration components and boundary condition based on Eley-Rideal kinetic equation and material balance principle were established while catalyst channel was chosen as research object. The equations were solved by fourth order Runge-Kutta method numerically. Considering calculation quantity and actual conditions the model was simplified and used to predict NOx reduction efficiency of 1#,2# units of JinQiao plant and 1# unit of XinFeng plant.The result revealed that the model can predict operating data of 1#,2# units both accurately, which demonstrated accurency and adaptability of mechanism model was better than black box mode such as neural networks. Compared to JinQiao power plant, XinFeng plant had poor prediction performance which revealed operating data must be tested by qualitative analysis model at first.The engineering application value of NOx reduction models had been evaluated while Jinqiao power plant was chosen as research object.The result demonstrated that range of operating parameters can identify abnormal situation during NOx reduction process. Analysis of correlation can identify abnormal between NOx production and ammonia flow. Quantitative model had good real-time and accurate prediction performance. Inlet NOx concentration model and NOx reduction efficiency model had better prediction performance and smaller relative error while NOx production and reduction model showed worse performance sometimes because of error stack. The model of outlet concentration model and NOx emission model was less affected and the prediction erro rwas within the acceptable range.