A Real Time Multi-Scale Correlation Analysis Method and its Application on the Data Mining in Thermodynamic Process

Abstract

There is a complicated mapping between the state and measurement signal in the complex thermodynamic system, direct mechanism analysis is very difficult. Using wavelet analysis idea for reference, simple filter bank was designed to decompose the signal by frequency domain. The correlation feature of signals in different frequency scale can be discovered. Using such method to analyze furnace pressure signal of boiler in thermal power plant, the positive correlation between the low-frequency component of furnace pressure signal and the medium-frequency component of fuel was revealed; and the negative correlation between the highfrequency component variance of furnace pressure signal and the low-frequency component of fuel was discovered. Through further mechanism analysis the conclusion is that: the furnace pressure can reflect the change of the furnace temperature in medium-frequency; the intensification of the furnace pressure and furnace temperature fluctuation in boiler under low load condition is cased by the nonlinearity of object itself.