Prediction Method Of The NO_x Emissions From Thermal Power Plant Based On Long Short-Term Memory Neural Network

Despite the continuous implementation of power structure reform in recent years,thermal power generation will remain the main form of power generation in China for a long time.The NOx produced in the combustion process of coal in a thermal power plant is one of the most important sources of air pollutants,which has a serious impact on human health and air quality.In response to the call of national energy saving and emission reduction to reduce environmental pollution,it is imperative for thermal power plants to control NOx emissions.In this paper,the combustion process parameters of 660 MW ultra-supercritical unit coal-fired boiler in a thermal power plant in Henan Province are taken as the research object.In view of the massive system data of thermal power plant,the multivariable mutual coupling characteristics of the coal-fired boiler and the DCS stores a dynamic,high-dimensional and massive time series data,which shows the interrelated characteristics between current and historical working conditions,time series forecasting method is introduced into NOx forecasting of thermal power plant in this paper.And a NOx emission forecasting model based on long short-term memory neural network is designed by using deep learning method.The main research contents are as follows:(1)Two sets of DCS historical data of a thermal power plant in Henan Province of different scales were collected as sample sets.Firstly,Z-score is standardized for input samples,and then principal component analysis is used to extract features,which effectively eliminates the interference of redundant feature variables on the prediction results,laying a foundation for the establishment of NOx prediction model.(2)The long short-term memory neural network which can dynamically memorize historical information is used to model,and the effects of the number of neurons in the hidden layer,the time step,the dropout rate and the optimization algorithm on the prediction results of the model are analyzed and compared.The optimal parameters of the model are determined by comparing the experimental methods.(3)The results show that the predictive effect of long short-term memory neural network is better than that of traditional recurrent neural network under the same model parameters.It shows that the designed long short-term memory neural network can effectively solve long term memory.In addition,the prediction of the model has less fluctuation and higher stability.(4)In order to ensure the interpretability of the data in the actual industrial field,the long short-term memory neural network model is improved,and the standardized data are directly input into the model.The predicted results are similar to the previous predicted results,which verifies that the long short-term memory neural network model is applied to the combustion of thermal power plants.The accuracy of large-scale high-dimensional data prediction of coal boilers lays a solid foundation for future combustion optimization work.The long short-term memory neural network constructed in this paper has achieved good prediction results on both small-scale data sets and large-scale data sets.It is superior to traditional methods in both prediction accuracy and training speed,and proves the reliability of the prediction model.The research results of this paper have certain reference significance for the application of in-depth learning in industrial field.