Vibration Trend Prediction of Pumped Storage Unit Based on VMD, SSA and Optimal KELM

Predicting the vibration trend of pumped storage units (PSU) plays an important role in the safe operation. For this purpose, a hybrid vibration trend prediction method of PSU based on VMD, SSA and optimal kernel extreme learning machine (KELM) is proposed. Firstly, VMD is used to decompose the collected vibration data of PSU into multiple subsequences. Then, SSA is employed to separate each subsequence into dominant part and remaining part, while the remaining parts of all subsequences are integrated to form a new subsequence. Later, phase space reconstruction (PSR) is applied to create samples for training KELM to achieve vibration trend prediction of PSU. When predicting with KELM, its parameters are determined with improved parrot optimizer (IPO). Finally, the prediction results of all subsequences are accumulated to obtain the ultimate value of vibration trend. Experimental results and comparative analysis ascertain the effect of the method.