Research On Data-driven Combustion Optimization Technologies For Coal-fired Boilers

Under the guidance of the Emission Peak and Carbon Neutrality policy,combining with the reality of China’s power development,coal-fired units are developing towards large capacity and ultra supercritical units.At the same time,we vigorously promote the new energy power generation system,which puts forward higher requirements for the flexibility of peak shaving and frequency regulation of coal-fired units,and faces the challenges of safe,stable and economic operation under low load and variable load conditions.Deep load regulation capacity is an inevitable requirement for thermal power units to adapt to the new situation,but the rapid and deep load regulation will have an impact on the safety,stability and economy of coal-fired boilers.The establishment of a prediction model for boiler furnace temperature and NOx emission can provide a foundation for rapid and deep load regulation of thermal power units.The study of multi-objective combustion optimization technology is of great significance to optimal control and efficient energy saving and emission reduction of thermal power units.This paper focuses on the research of combustion optimization technologies for coal-fired boilers.The main research contents are as follows:1.Through the heat and mass transfer principle and model analysis of the combustion process in the main combustion zone of boiler furnace,a new infrared temperature measuring device is developed to obtain the temperature of the combustion layer at different heights in the main combustion zone online and in real time for a long time.Based on Pearson correlation coefficient,the relationship between combustion layer temperature and boiler load,main steam temperature and pressure,NOx concentration and lower load burns steadily is explored.This paper analyzes the important role of combustion layer temperature on the safe,stable,economical and efficient operation of coal-fired boilers,and provides reference for the realization of deep load adjustment,low load burns steadily and low nitrogen emission of coal-fired boilers.2.Considering that combustion layer temperature plays an important role in coal-fired boilers,an online identification and prediction model of combustion layer temperature is proposed based on entropy fuzzy clustering algorithm.Combined with the historical operation data of coal-fired boilers,an online identification model of combustion layer temperature is established based on fuzzy clustering and subspace identification model.The results show that the identification model has the characteristics of strong interpretation and easy implementation,and provides the possibility for combustion optimization of coal-fired boilers.3.In view of the characteristics of coal-fired boilers,such as multi-variable coupling,nonlinearity,instability and large lag,an Complementary Ensemble Mode Decomposition with Adaptive Noise(CEEMDAN)algorithm is proposed by studying the empirical mode decomposition algorithm.Through CEEMDAN decomposing the key operating parameters and NOx emission sequence of coal-fired boilers,the factors affecting NOx emissions of coal-fired boilers are analyzed and studied,which provides a basis for the prediction modeling of NOx emissions in coal-fired boilers.4.In view of the influence of large load adjustment of coal-fired boilers on NOx emissions,combined with the characteristics of non-linear and unstable NOx concentration sequence,this paper uses CEEMDAN algorithm to decompose the original NOx into a series of intrinsic mode functions(IMFs)to extract the data features.On this basis,Attention Mechanism(AM)and Long Short-Term Memory(LSTM)algorithm are used to study the dynamic prediction modeling of NOx concentration under different load fluctuation conditions.The results show that the NOx concentration prediction model based on CEEMDAN-AM-LSTM can accurately predict the NOx concentration under full load conditions,and provide support for boiler load depth adjustment and low nitrogen combustion optimization.5.In view of the characteristics of deep load adjustment and lower load burns steadily of coal-fired boilers,the boiler efficiency model is established based on the inverse balance method by comprehensively considering the complex nonlinear and mutual coupling relationship among the multiple input operating parameters.Meanwhile,the paper studies the dynamic models of NOx concentration and boiler efficiency based on Gaussian process model under different load conditions of coal-fired boilers.Then,the improved non-dominated sorting genetic algorithm(NSGA-Ⅱ)was used to carry out multi-objective optimization of boiler efficiency and NOx concentration,and the optimal boiler operation parameters were obtained to meet the NOx concentration index,the results provides the decision-making reference for low nitrogen and high efficiency operation of coal-fired boilers.