Vibration Prediction of Hydropower Unit Based on Adaptive Multivariate Variational Mode Decomposition and BiLSTM.

Vibration trend prediction of hydropower units is of great significance to ensure the safe and stable operation of hydropower units. To address the limitations of existing models for predicting the vibration trend of hydropower units. In this paper, we propose a combined trend prediction model based on adaptive multivariate variational mode decomposition (WOA-MVMD) and bidirectional short-duration memory neural network (BiLSTM). The model adopts multivariate variational modal decomposition (MVMD) to decompose multi-channel data synchronously, retains the coupling between the original data channels, adopts whale optimization algorithm (WOA) to optimize the selection of MVMD decomposition parameters, avoids the shortcomings caused by manual parameter selection, and realizes the optimal adaptive decomposition of vibration sequences. A series of IMF sub-sequences obtained from modal decomposition are normalized. Then the BiLSTM trend prediction network is established for each subsequence signal, and the final prediction result is obtained by superposition and reconstruction of the subsequence prediction results. Based on the actual operation data of a power station in China, the proposed model is proved and tested, and the high prediction accuracy of the proposed model has been verified. [ABSTRACT FROM AUTHOR]