Aiming at the characteristics that the traditional Prony algorithm is easily interfered by noise and there is correlation between multiple power quality signals in the same area, in this paper, the Prony harmonic detection algorithm based on multi-channel signal joint denoising is proposed to achieve the accurate detection of harmonics under strong noise conditions. Firstly, the central frequency method and trajectory similarity method are used to improve the multivariate variational mode decomposition algorithm.Then, the improved MVMD algorithm is used to jointly decompose the associated multi-channel signals, extract the dominant mode components, and reorganize them into stable signals suitable for Prony harmonic analysis. Finally, Prony analysis is performed on the stable signals to obtain preliminary harmonic parameters, and the threshold screening and artificial fish swarm global optimization are carried out to obtain the accurate harmonic detection parameters. Simulation experiments show that the output signal-noise ratio of the improved MVMD denoising algorithm is 37.3, which is higher than VMD denoising method (33.2) and wavelet denoising method (32.8),and the denoising effect is better; The error of the harmonic detection result of the algorithm in this paper is generally less than that of the traditional Prony algorithm. It possesses the characteristics of high harmonic detection accuracy and simultaneous calculation of multi-channel signals.