Cavitation detection via motor current signal analysis for a centrifugal pump in the pumped storage pump station

Cavitation is quite common during centrifugal pump operation which degrades the safety and stability of the pumped storage power station. Instant prognostication of incipient cavitation and precise status monitoring of cavitation evolution can benefit accuracy of cavitation detection. In this research motor current signal analysis (MCSA) technique is applied for cavitation quantitative characterization. In order to improve the performance of MCSA for cavitation detection, the method of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is used for noise elimination and feature selection. IMFs irrelevant with cavitation are removed based on cavitation numerical simulation and the concurrent frequency bands in spectrograms drawn with CEEMDAN. According to the feature analysis for incipient cavitation, the signal power of IMF 3–9 is extracted to reveal incipient cavitation. The peak value in the marginal spectrum of IMF 7–8 is extracted as the indicator for the characterization of cavitation evolution. Both indicators show faster and more precise performance than previous work and can be suitable for larger-scale working conditions. Thus, CEEMDAN is beneficial for improving the feasibility and accuracy of MCSA technology. This research provides technical assistance for cavitation detection and prevention in the pumped storage power station. Failure Detection; Centrifugal Pump; Motor current signal analysis (MCSA); Cavitation characterization; Centrifugal pump; Indicator extraction.