Vibration Tendency Prediction of Hydroelectric Generator Unit Based on Fast Ensemble Empirical Mode Decomposition and Kernel Extreme Learning Machine with Parameters Optimization

Abstract

Since the existence and effect of non-linear and non-stationary characteristics for hydroelectric generator unit (HGU) vibration, an intelligence vibration tendency prediction method based on fast ensemble empirical mode decomposition and kernel extreme learning machine (FEEMD-KELM) is proposed to obtain better prediction results. Firstly, the vibration signal is decomposed into several intrinsic mode functions (IMFs) by FEEMD. Then, the predict models of KELM are constructed. Meanwhile, the salp swarm algorithm (SSA) is used to identify the number of hidden layer nodes of each KELM model. Finally, all KELM predictions are summed to obtain the predicted values of the original vibration signal. A case study of the mixed-flow hydropower unit vibration data in China is carried out, and the experimental results demonstrate that the proposed method can achieve better predictions in practical applications.