This study proposes a high-precisionwavelength demodulation algorithm of fiber Bragg grating (FBG) based on wavelet denoising to overcome the shortcomings of traditional wavelength demodulation algorithms in the online monitoring system of FBG, such as low detection accuracy and effectiveness. First, we perform rapid calculations on the spectral waveforms that are processed using the wavelet noise reduction method to generate a new function. Then, we obtain the center wavelength for the new function. Finally, a set of FBG transformer winding temperature online monitoring system based on a vertical-cavity surface-emitting laser is constructed; the wavelet noise reduction technique is used to demodulate the wavelength and indirectly acquire temperature data. The experiments show that the wavelet denoising technique outperforms the finding peaks directly, polynomial fitting, and Gaussian fitting algorithms in terms of speed and stability. In the temperature range of 70-90 ℃, the temperature measurement accuracy can reach ±0.04 ℃. The wavelength demodulation's highest standard deviation is less than 1.67 pm. Further, the operational efficiency of the algorithm is greater than that of the commonly used Gaussian fitting algorithm. Therefore, the algorithm may be used in the FBG's high-precision real-time online monitoring system in the industrial environment.