Research On NO_x Prediction And Coal Blending Optimization Of Power Station Boiler Based On Data Mining

Coal-fired power generation is the main power generation structure in China,and it will remain the main form of power generation for a long period of time.Due to the large changes in the coal market and the pursuit of low cost by power plants,the coal blending method is usually selected for unit operation.However,the current emission requirements are becoming more stringent and coal types are more complex and diverse,the traditional method of coal blending based on manual experience has many drawbacks,such as low decision-making efficiency and high operating costs due to misjudgment.At present,artificial intelligence technologies such as machine learning have developed rapidly and have been applied in many fields.In the field of power generation,prediction models for NO_x and other indicators have a lot of research in the fields of combustion optimization and online monitoring,but less research on coal blending.At present,coal blending is still mainly based on manual experience or expert systems.Thermal power station is a scene with massive historical data,through deep mining of the massive data of thermal power station,the establishment of prediction models aimed at guiding the coal blending is of great significance for assisting manual decision-making.The main considerations when blending coal are NO_x emissions,boiler efficiency and other indicators.The article firstly takes NO_x emissions as the research object,determines the initial variables through the analysis of NO_x generation mechanism the data preprocessing,finally establishes the boiler NO_x emission prediction model based on LightGradient Boosting Machine(LightGBM)through the processes of data preprocessing,feature engineering and modeling tuning,which uses the Random Search method and Bayesian Optimization to optimize the hyperparameters.The performance of the model is verified by comparison between models and comparison before and after optimization,the results show that the performance of the NO_x prediction model based on LightGBM is better than the BP neural network model,support vector machine model,random forest model and XGBoost model,verifying the advanced nature and accuracy of the model established in this paper.At the same time,with the help of optimization algorithms,the accuracy of the model established in this paper is further improved.Secondly,the LightGBM model is applied to the prediction of exhaust gas temperature,the result shows that the model also has high accuracy and excellent generalization ability.According to the analysis of actual application scenarios of blending coal blending,the established NO_x prediction model and exhaust gas temperature model are simplified.The Pearson correlation coefficient is used to determine the operating parameters that are strongly linearly related to the load,and the regression equation is obtained by the method of least squares linear regression.The result shows that the simplified NO_x prediction model and the exhaust gas temperature prediction model maintain high performance,verifying that the prediction models have practical application value.Finally,the application method and output result of the predictive models are introduced through a case,and the predicted value of the model can be used as an important basis to guide fuel operation and has considerable engineering value.