Quantitative Analysis Of Furnace Flame Stability Based On Fire Detection Signal

The correct detection and evaluation of furnace flame stability of coal-fired power station boilers plays an important role in the safe and economic operation of boilers.During the operation of the pulverized coal furnace,the unstable combustion of pulverized coal in the furnace may cause a series of problems that threaten safe and economic operation.Especially in recent years,the burning tendency of low-quality coal and blended coal has increased,leading to increased instability of combustion.Therefore,it is particularly important for the operation of the boiler to accurately,stably and reliably detect the combustion state of the flame in the furnace and prevent the occurrence of abnormal situations during the operation.This paper takes a 660MW unit in a power plant as an example.Based on the application of the ABB fire detection analysis unit,using the collected flame detection intermediate values,with the help of cluster analysis,a qualitative judgment method for flame stability is obtained.By comparing the clustering results of the physical quantities in the furnace flame combustion in the actual power plant operation,combined with the actual operation situation,analyze the limitations of the K-Means clustering(K-Means),Density Clustering(DBSCAN)and the Balanced Iterative Reducing and Clustering using Hierarchies(BIRCH)when used for the actual combustion detection data.Finally,with the help of hierarchical clustering agglomerated(HCA),the state division process of the clustering tree diagram is analyzed,combined with the change of coal supply,the clustering tree is re-divided,and a furnace combustion state classification method consistent with actual operating experience is obtained.Based on the hierarchical clustering to divide the combustion state,the combustion quality index reflecting the stability of the furnace flame is defined.The value of this index is between 0 and 100.The value of the index reflects whether the combustion state is stable.The closer the value is to 100,the more stable the combustion is;the closer to 0,the more the combustion tends to the extinguish state.Finally,the rationality of the combustion quality index is analyzed based on the change of coal feed rate.After obtaining the classification of the combustion state through the HCA method,in order to further explore how to use the intermediate value of the fire detection to describe other forms of the combustion state of the furnace,the short-time Fourier transform is used to obtain the time-frequency diagram of the intermediate value of the fire detection,and established a Convolutional Neural Network(CNN)model for flame stability discrimination,trained the model with time-frequency diagram of a given state label,finally verified the recognition of the test dataset by the network,and achieved a high accuracy rate.