Multi-step vibration trend prediction model based on multi-objective optimization for hydropower generator unit

Abstract

Hydropower generator unit (HGU) is the vital equipment in frequency and peaking modulation of the power grid, whose vibration contains a wealth of status characteristics. Therefore, it is significant to predict the vibration tendency of HGU and it is helpful to achieve predictive maintenance as well. However, most prediction models only focused on enhancing the stability or accuracy in previous studies. In this paper, an intelligence vibration tendency multi-step prediction model is proposed to achieve simultaneously strong stability and high accuracy, which is based on variational mode decomposition (VMD), multi-objective salp swarm algorithm (MOSSA) and kernel extreme learning machine (KELM). Firstly, the vibration signal is decomposed into several modes by VMD. Then, the prediction models of KELM are constructed. Meanwhile, MOSSA is used to identify the parameters in each KELM model. Finally, all KELM predictions are summed to obtain the predicted values of the original vibration signal. To investigate the multistep prediction performance of the proposed model, a case study and analysis of the mixed-flow HGU data in China is carried out. The experimental results demonstrate that the proposed model can achieve better results in terms of predicting stability and accuracy.