Research On Soft Measure Of Carbon Content In Coal Ash Based On Data Driven

In the real-time process of boiler running,the carbon content in coal ash directly reflect the utilization of coal and is also an important indicator of boiler efficiency.In the running process to most of medium and small boiler,due to limited by measuring means,only through the weightlessness method of online combustion to measure the value of carbon content in coal ash,the method is time-consuming and large-lag.It is difficult to adjust the process of boiler combustion by the parameter.Soft measurement technology is one of the effective ways to solve this problem with the advantages of strong real-time,and variety of modeling methods.Boiler combustion is a complex industrial process,there are relationships of nonlinear,strongly coupled.First designing the common modeling method-multiple stepwise regression.Though this way,we can get a kind of linear mathematical expression,which could determine the structure and parameters of model.The relationship between the various variables of the regression model and the carbon content in coal ash is linearized and the regression mode can basically reflect the change trend of carbon content in coal ash in the specific section.When the carbon content in coal ash is larger or smaller,duo to the generalization ability of the model is poor,it could not reflect the actual value of carbon content in coal ash.So this linearized approach is not appropriate.The neural network has the ability of dealing with the nonlinear problem and strong self-learning,which can use the different weights to express the weight of process parameters.But people are not involved for this modeling process.The fuzzy system can make full use of experience and knowledge and the process of modeling is no longer a "black box"process.Between Neural network and Fuzzy System has obvious complementary advantages.Adaptive Neural Fuzzy Inference System(ANFIS)formed by a combination of two has both advantages.Establishing the model of carbon content in coal ash by ANFIS,selecting different variables and partitions and different membership functions,thus produce a comprehensive rule library.Adjusting the function of membership by mechanism analysis,in this way make the precision of the model is more widely applied.The average error rate of the ANFIS model is 2.22%lower than the neural network model,which is 3.06%lower than the multivariate stepwise regression model.Therefore,the ANFIS model is more accurate and more suitable for soft measurement of carbon content in coal ash.Apple the model of carbon content in coal ash established by ANFIS to a university's DH29-1.6/130/70-AII boiler.According to the value of the carbon in coal ash in the online real-time soft measurements optimize the process of boiler combustion,the carbon content in coal ash is down 2.45%on average,increasing the utilization rate of coal.The soft measure of carbon content in coal ash can provide positive guidance for the combustion of boiler.