基于VMD的抽水蓄能机组振动参数演化预测

提出了一种基于移动最小二乘响应面和变分模态分解（Variational mode decomposition，VMD）的抽水蓄能机组振动参数演化预测方法。首先利用移动最小二乘响应面建立抽水蓄能机组振动参数实时评估模型。然后利用VMD将复杂非线性的机组振动参数时间序列分解若干个平稳分量时间序列。其次对每个分量进行特性识别，根据其不同属性，分别采用LS-SVM或GM（1，1）对每个分量进行预测。最后重构每个分量的预测值获得原始时间序列最终的预测结果。实例分析表明，该方法能较准确地预测机组振动参数演化趋势。

Vibration parameter development prediction of pumped storage unit based on VMD

A vibration parameter development prediction method of pumped storage unit based on moving least square response surface (MLSRS) and variational mode decomposition (VMD) is proposed. The ML-SRS is adopted to construct health condition assessment model of pumped storage unit vibration parameter. Then VMD method is applied to decompose nonlinear vibration parameter time series of unit into several stable components. According to different features of each component, LS-SVM or GM (1, 1) is selected to predict them respectively. Finally, predicted results of each component are added up to obtain the pre-dicted results of original vibration parameter time series. The results of real vibration data show that the pre-diction model is effective and of high-accuracy.