Abstract: Against the backdrop of the strategic goal of "peaking carbon emissions and achieving carbon neutrality" proposed in the national "14th Five Year Plan", hydropower new energy has also ushered in new opportunities and challenges for the industry. With the expansion of power stations, the traditional manual inspection combined with industrial monitoring often faces problems such as inability to automatically identify and judge faults, and low sensitivity to information feedback. By combining 5G technology and artificial intelligence, variational mode decomposition and image grayscale processing techniques were introduced to analyze the operating data of power plant units. The results showed that the fractal dimension of the image has frequency characteristics of 30 Hz and 85 Hz, with amplitudes corresponding to 0.02 and 0.009, respectively, which are the main frequency and secondary frequency, and far stronger than other clutter frequencies. The VMD method successfully decomposed the pressure pulsation signals of each monitoring point and obtained the characteristics of various modal functions in the time and frequency domains. By analyzing the VMD decomposition results of two monitoring points at the tail water pipe, it was found that their frequency components were similar and consistent with the frequency of the monitoring points inside the volute. The research results of this article can provide important support for the intelligent construction of power stations, and also provide more effective means for the operation and maintenance of power stations.