The task of flux calibration for LAMOST (Large sky Area Multi-Object Spectroscopic Telescope) spectra is difficult due to many factors. For example, the lack of standard stars, flat fielding for large field of view, and variation of reddening between different stars especially at low galactic latitudes etc. Poor selection, bad spectral quality, or extinction uncertainty of standared stars not only might induce errors to the calculated spectral response curve (SRC), but also might lead to failures in producing final 1D spectra. In this paper, we inspected spectra with Galactic latitude |b| ≥ 60• and reliable stellar parameters, determined through the LAMOST Stellar Parameter Pipeline (LASP), to study the stability of the spectrograph. To guarantee the selected stars had been observed by each fiber, we selected 37,931 high quality exposures of 29,000 stars from LAMOST DR2, and more than 7 exposures for each fiber. We calculated the SRCs for each fiber for each exposure, and calculated the statistics of SRCs for spectrographs with both the fiber variations and time variations. The result shows that the average response curve of each spectrograph (henceforth ASPSRC) is relatively stable with statistical errors ≤ 10%. From the comparison between each ASP-SRC and the SRCs for the same spectrograph obtained by 2D pipeline, we find that the ASPSRCs are good enough to use for the calibration. The ASPSRCs have been applied to spectra which were abandoned by LAMOST 2D pipeline due to the lack of standard stars, increasing the number of LAMOST spectra by 52,181 in DR2. Comparing those same targets with SDSS, the relative flux differences between SDSS spectra and that of LAMOST spectra with the ASPSRC method are less than 10%, which underlines that the ASPSRC method is feasible for LAMOST flux calibration.-3 -