In order to realize the real-time monitoring of the reflection spectrum signal in the fiber Bragg grating (FBG), the hardware system and the software architecture for demodulating the FBG signals were designed based on the all-solid-state spectrum module. The host computer software was programmed and embedded with different peak-detection algorithms, which could be applied to the calculation of the central wavelength under different noise environments to reduce the wavelength error caused by the algorithm. The software also included functions such as data acquisition, storage and playback. Taking Gaussian fitting peak-detection algorithm as an example, a method of simulation evaluation to peak-detection algorithm without hardware device was proposed, which improved the development efficiency of this kind of spectrum demodulation system. The simulation evaluation experiments were performed by using the ideal FBG reflection spectrum with different noise levels. The results show that the best peak-detection result can be obtained when the threshold is taken as twice the noise average amplitude, and the peak-detection errors are inversely proportional to the signal-to-noise ratio (SNR) of the collected data and the spectral peak width. The Gaussian fitting peak-detection algorithm is used to calculate the FBG reflection spectrum data actually collected by the hardware system, and the obtained accuracy of the system is 5.89 pm, the error from the simulation evaluation result is less than 1.7 pm, which proves the feasibility of the simulation evaluation.