In the process of synchronous condenser fault diagnosis, the impact of noise signals can easily lead to poor diagnostic accuracy and increase the burden on power grid operation. To this end, a fault diagnosis method for synchronous condensers in ultra-high voltage transmission systems was proposed. By analyzing the vibration of the adjusting camera bearing machinery, the vibration signal generated by the abnormal vibration of the camera bearing machinery was collected, and the signal noise was eliminated by the ensemble empirical mode decomposition, and the optimized signal with smoother waveform was obtained. The vibration signal characteristics were obtained by calculating the energy entropy corresponding to the center frequency of the signal modal component, and the signal characteristics were input into the support vector machine as training samples to achieve fault diagnosis of synchronous condensers for ultra-high voltage transmission systems. The experimental results show that the diagnostic results of the proposed method for the fault points α, β, and γ coincide with the actual waveforms of current and voltage at the corresponding points. The coincidence rate of the harmonic current amplitude detected at the short-circuit points in groups a, b, and c and the actual harmonic current amplitude at each peak is relatively high. It shows that the proposed method has good diagnostic performance of open circuit fault and short circuit fault, which is beneficial to ensure the stable operation of UHV transmission system.