Yes, I did use "wavelength sweep". As for your advice, I set EME solver with different wavelength and then do wavelength sweep. Here are the results of different center wavelength.

Which result is resonable? I think that every independent center wavelength is resonable, rather than the sweep result of a series of wavelength. Besides, the dB scale is changed with the percent scale and the ripple is still serious, for an edge coupler.
k (in n,k) of SiN is 0, so we do not consider dispersion. The boundary conditions and the input source are the same with the example. Hope for your reply,