Abstract: To improve the safe and stable operation of pumped storage units, this paper analyzes their operational status, obtains the health condition of unit equipment, and accurately predicts their future development trend. It proposes a trend prediction model that integrates variational mode decomposition（VMD） and attention mechanism（AM） into a bidirectional long short-term memory network（BiLSTM）. Firstly, a healthy state model of the unit is established by using the Bagging algorithm, which considers the influence of factors such as active power, working head, guide vane opening, and speed. Secondly, based on the healthy state model, the degradation trend sequence of the unit is calculated, and the VMD algorithm is used to decompose the trend sequence into multiple smooth and stable modal components. Finally, a BiLSTM network and attention mechanism model is established for each modal component to predict the trend, and the predicted results of each component are superimposed to obtain the final trend prediction result of the unit. Simulation results show that the proposed method can accurately express the degradation trend of the unit and effectively improve the prediction accuracy of the degradation trend.