Boiler Combustion System Economic Analysis Based On Big Data Mining With Multi-Objective Optimization

With the development of science and technology and the rise of smart power plants,the DCS of thermal power plants stores massive data,and the use of big data mining technology to optimize the operation of thermal power units is of extraordinary significance for improving the economy and environmental protection of thermal power plants and promoting China’s low-carbon energy transformation.Based on the historical operation big data of a thermal power plant,this paper uses data mining technology to deeply mine the optimal information contained in the 350 MW thermal power unit of a power plant,and uses multi-objective optimization algorithm to carry out prediction model and optimization analysis of boiler combustion system,so as to reduce NOx concentration and improve boiler efficiency.Firstly,the influencing factors of boiler efficiency and NOx emission were systematically analyzed,and then,taking the historical data of a 350 MW unit of a power plant in Jilin as the research object,the boiler efficiency was calculated by the inverse balance method,and the data preprocessing of the big data of the unit was carried out,including: outlier value removal and missing value filling,using the sliding window method to extract steady-state data,maximum information coefficient feature selection algorithm to exclude irrelevant data,reduce the number of input dimensions,etc.,to pave the way for subsequent economic analysis and model establishment.Secondly,using big data mining technology to deeply explore the optimal values of boiler controllable operating parameters and boiler thermal efficiency,in view of the limitations of the traditional K-means algorithm,a DC-K-means improved by using density thinking and center point substitution method is proposed,DC-K-means algorithm is used to divide and discretize working conditions according to external constraints,and finally Apriori is used to mine the potential correlation contained in boiler combustion.The correlation rule analysis can obtain the optimal value of the parameters under the whole working condition and obtain the boiler combustion rule.It can effectively improve boiler efficiency,reduce coal consumption and improve economic benefits.Thirdly,ELM is used to model and predict boiler efficiency and NOx concentration,MIC algorithm is used to screen the model auxiliary variables,and the information coefficient MI is used to calculate the delay time of the auxiliary variables and reconstruct the modeling data.Then,the NOx prediction model was established by using the ELM,and the surface model was accurate and effective after comparison with other models.Then,a boiler efficiency prediction model is constructed based on the ELM model,and the experimental results verify the accuracy and effectiveness of the model.Finally,based on the NOx and boiler efficiency prediction model,a multi-objective optimization model with boiler thermal efficiency and NOx as the objective function is constructed.Specifically,it includes a two-stage decision-making method: the first stage uses NSGA-Ⅲ to obtain the Pareto optimal solution set,and compares the convergence,distribution and stability of the algorithm;In the second stage,a multi-attribute decision-making method based on FCM-GRP and the only optimal solution is selected from the Pareto optimal solution set.Finally,the multiobjective optimization model of boiler combustion is applied to the actual operation unit,which achieve "high efficiency and low emission".The optimized value of the decision parameters obtained has an important reference value for the economic and environmental protection operation of coal-fired power station boilers.