The Application Research Of Frequency Domain Identification In The Thermal System

Frequency responses reflect dynamic thermal characteristics more accurately. The frequency domain identification, one of the important methods in complex thermal system identification, can get the transfer function parameters through its frequency response. This thesis proposes a separated-frequency identification algorithm based on least-squares principle and wavelet analysis theory. With less computation the algorithm also has the advantages of both frequence domain and time domain identification. By adjusting the weights the identification method can remove high frequency noise and highlight the low frequency characteristics. Simulations show the validity of the algorithm. As this algorithm will get the different results when wavelet changed the article takes Daubechies wavelet as example, analyzes the impact of vanishing moment in the separated-frequency identification and takes this as the basis for the choice of db wavelet. Further more a better decomposition scale is selected by comparing the errors of identification results. Simulations of typical thermal instance identification show the algorithm can achieve a good performance with appropriate wavelet and scale.