NO_x Emission Prediction Of Coal Fired Boiler Based On Improved Hybrid Kernel Extreme Learning Machine

In recent years,with the further expansion and opening of China’s power market and the increasingly stringent requirements of the national environmental protection policy,thermal power enterprises have optimized and upgraded the coal-fired power station boilers with "low economic efficiency and high pollution",so as to achieve the dual goals of high economic efficiency and low pollution emissions of thermal power enterprises.The rapid development of big data analysis technology,machine learning and other artificial intelligence technology provides theoretical and technical support for the optimization and upgrading of coal-fired power plants.The optimization effect depends on whether the NO_x emission can be accurately predicted.At present,how to accurately predict NO_x emission and scientifically evaluate NO_x emission prediction effect is still the focus of research.In this study,firstly,the historical operation data of boiler are preprocessing,including error data and missing data processing,re-sampling,normalization and so on;Then,on this basis,we mainly carry out the following three aspects of research:(1)In view of the problem that the data redundancy of coal-fired power plant boiler affects the accuracy of NO_x prediction,a FAR-HKELM prediction model based on FAR and HKELM is proposed by fully considering the correlation between input variables and output variables.Based on actual historical operation data,this paper compares the modeling methods of HKELM,FAR-HKELM with BP,SVM,ELM,PKELM and GKELM,and verifies the effectiveness and superiority of FAR-HKELM.(2)Aiming at the error distribution problem of non-Gaussian characteristic data of coal-fired power plant boiler,this paper uses the local similarity function MCC as the performance evaluation index,and a MCCE-HKELM prediction algorithm based on MCCE and HKELM is proposed,that is,a NO_x emission prediction method of coal-fired boiler.Firstly,the NO_x prediction model based on HKELM is established,and then the local similarity function is used as the performance evaluation index,and the particle swarm optimization algorithm is used to find the optimal parameters of the model with the goal of maximizing the local similarity function,so as to improve the performance of the prediction model;in addition,the relevant comparative experiments are set for the two selection methods of MCC kernel function parameter value σ,so as to provide the reference for the kernel function The selection of value σ provides theoretical guidance.(3)Aiming at the problem that data noise and outliers affect the performance of NO_x prediction model for coal-fired power plant boiler,and considering that HKELM is derived under the minimum mean square error criterion of Gaussian noise assumption,the performance of model based on HKELM will deteriorate seriously in the case of non-Gaussian noise.In order to improve the robustness of HKELM,a prediction algorithm of MMCC-HKELM based on MMCC and HKELM is proposed,and its regression prediction performance is verified on the Artificial simulation data set and six benchmark data sets;In addition,the NO_x emission prediction model of coal-fired boiler is established based on MMCC-HKELM algorithm,and experiments are set to analyze the model,which verifies the performance advantage of this method in NO_x prediction of coal-fired power plant boiler.