Position shaking of reference center wavelength is the main factor which affects the accuracy of absolute wavelength calibration. In this case, peak-detection algorithms are proposed to find the real-time position of center wavelength. Five peak-detection algorithms are analyzed and compared by simulations and experiments. The error of Gaussian nonlinear curve fitting algorithm is 0.15 pm, which is the lowest in all the test algorithms. The influence of power threshold value upon the five peak-detection algorithms is studied, and the importance of threshold optimization is clear and definite for reducing peak-detection errors. After threshold value optimization, the Gaussian nonlinear curve fitting algorithm performs the best, as its lowest error is 0.04 pm and the average error is 0.06 pm. Thus, Gaussian nonlinear curve fitting algorithm meets the requirement of calibration accuracy. By analyzing factors which cause error in the peak-detection algorithms, we find that the signal noise ratio turns out to be the major factor which dominates the errors of Gaussian nonlinear curve fitting algorithm. Therefore, noise suppression is the best way to achieve high overall accuracy of absolute wavelength calibration.