Study Of Analysis Methods For Thermal Signal Based On Wavelet Transform

The stability of boiler combustion directly affects the safety and economy of the unit. Using SIS platform data, the research is focus on this two aspects:the relevance of the thermal signal, the extraction of combustion-related signal. In order to overcome the insufficiency of correlation information for overall tendency and local fluctuation in the analysis of traditional thermal signals,a new wavelet multi-resolution strategy with correlation in different frequency ranges was proposed.Firstly,components of different frequency ranges were obtained via multi-level wavelet decomposition and reconstruction.Then,correlation coefficients were calculated based on the assumption of fluctuation similarity among components in the same frequency range,which can be used to describe the correlation degree of those components.The proposed method was utilized in the data analysis procedure of a 600 MW thermal unit.The result shows that the correlation level of frequency tendency in a low frequency range can be calculated by quantitative analysis;and signals with strong similarity in high frequency range can be mined.Thus,mining ranges of related thermal signals are widened by the proposed method. As in the same frequency band, disturbance-related signals have a strong similarity of volatility, a method using wavelet entropy has been presented to extract correlated signals of combustion disturbance. Components of different frequency bands were firstly filtered using wavelet transform, then calculating the wavelet entropy to analyze the correlation between signals. In addition, with the signal of frequency-dependent as the object, the influence of the wavelet basis and sliding window width on the identification performance of the wavelet entropy is discussed deeply. An example analysis based on a power plant show that this method can effectively excavate correlated signals of combustion disturbance, and can distinguish the relative strength.