Thermal process robust identification using wavelet de-noise and least-squares method

Unmodeled dynamics exit when modeling with transfer function model set for thermal process. Unmodeled dynamics and stochastic disturbances are considered as colored noise which makes least-squares estimates bias. De-colored noise using wavelet MRA is beneficial to enhance identification accuracy and robust. This is also proved by simulation with a linear time-invariant system. The proposed method processed operation data comes from thermal process and got satisfying result.