Pattern Recognition Of Combustion State Of Thermal Power Units Based On Operational Data

In recent years,the transformation of my Chinese energy architecture and the proposal of dual-carbon target have brought new challenges to the development of thermal power unit.It is urgent for thermal power unit to improve their flexibility and better meet market demands.With the advent of the Internet and big data era,the advanced technologies represented by artificial intelligence technology has been widely used in the power industry,improving the level of automation,informatization and digitization of thermal power plant,and promoting the construction of intelligent power plants.The combustion optimization technology based on big data analysis has become a new way to achieve safe,economical,environmentally friendly and stable operation of thermal power unit.Based on the historical operating data of the power plant,this paper makes the following research on how to select the parameters related to the combustion state,identify the combustion state mode,and establish the typical sample library of the combustion mode:Firstly,it is studied that the selection of parameters to characterize the combustion state.Due to the nonlinearity and high dimensionality of the power plant operating parameters,the multi-scale correlation analysis method is adopted in view of the fact that the classical correlation analysis method can not directly obtain the correlation characteristics among the thermal parameters.According to the different fluctuation similarity of the operation parameters in different frequency bands,the operation parameters are decomposed into different frequency components by using the frequency decomposition characteristics of wavelet packet transform.By comparing and analyzing the correlation of many operation parameters in different frequency bands,the total fuel quantity,furnace negative pressure,feed water flow rate,reheat steam pressure and flue gas temperature at the inlet of air preheater are selected as signals to characterize the combustion state.Secondly,it is studied the division of boiler operating conditions and the selection of typical operating conditions.Since the typical sample library of multiple combustion modes is designed based on the matching of operating conditions,the operating conditions of the boiler are divided according to the start-stop status of the coal mill and the actual operating regulations of the boiler.On this basis,in order to reduce the complexity of the establishment of typical sample database and reduce the subjective factors of typical working condition selection,the typical working condition selection method based on moving least square method is proposed.By calculating the error between the fitting data of random operating conditions and the original data,combined with the experimental cost of combustion adjustment,13 typical operating conditions are selected from 26 operating conditions.Then,it is studied that the pattern recognition of combustion state and the establishment of sample database.In the steady-state range of typical operating conditions,the combustion state pattern is identified according to different optimization objectives.For identifying the comprehensive optimal mode,the economic index function is constructed by using the relationship between boiler efficiency and power generation coal consumption,denitrification inlet NOx concentration and urea flow rate.For identifying the optimal mode of deep peak regulation,a multi-level fuzzy comprehensive evaluation method based on state parameters is proposed.Through the construction of a comprehensive evaluation model of state parameters,the evaluation values characterizing the low-load combustion stability of the boiler are obtained,and the evaluation values are compared to identify the optimal mode of deep peak regulation.In different modes,the operation rules of operation parameters such as throttle opening and oxygen content are mined to obtain the best operation parameters of the corresponding state parameters.through the integration of parameter samples in different modes to complete the establishment of multi-combustion mode typical sample library.Finally,the research results of this paper are applied to the intelligent combustion control system of 660 MW unit in Shaanxi JJ Power Plant.Through the distributed control system,the configurations of working condition division,combustion mode division and typical sample database are built and the system interface is designed.