As the core equipment of hydropower plants, the healthy condition of hydropower generating unit (HGU) plays a vital role in the safe and stable operation of hydropower plants. Therefore, it is of great significance to measure the vibration trend of HGU, which can effectively reflect the health condition of HGU, allowing the development of appropriate countermeasures to improve the safety and stability operation of HGU. Given this, a hybrid approach for measuring vibration signals of HGU coupled with complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), phase space reconstruction (PSR), kernel extreme learning machine (KELM) optimized by hybrid slime mold algorithm and Harris hawks optimization (HSMAHHO), and error correction with gate recurrent unit (GRU) network is proposed in this paper. Specifically, CEEMDAN is initially applied to decompose the raw vibration signals into several intrinsic mode functions (IMFs). Subsequently, PSR is adopted to convert each IMF into the input–output matrix of KELM for prediction. Meanwhile, HSMAHHO algorithm is utilized to optimize the critical parameters within KELM. Afterward, the predicted values of each IMF are superposed to obtain the predicted values of the raw vibration signals, and the error sequence to be corrected is constructed. Eventually, the error sequence is predicted by combining CEEMDAN, PSR, GRU and then summed up with the previous predicted values to get the final measuring result. In addition, the feasibility of the proposed hybrid approach is further verified by the experimental comparative analysis with seven comparative models. The experimental results demonstrate that (1) the proposed HSMAHHO algorithm could better optimize the internal parameters of KELM, which effectively improves the measuring results (2) the proposed error correction strategy could effectively enhance the measuring accuracy.