条件驱动的非线性协整与非平稳状态监测

Condition-driven Nonlinear Cointegration and Nonstationary Operation Monitoring

针对工业生产过程中,非平稳变量间服从非线性协整关系所导致的非平稳特征提取难题,提出一种条件驱动的非线性协整算法。首先,根据设备的运行状态,对数据进行重构,将数据从时间轴重构至条件轴,使得每一个条件片内的数据具有相似的过程特性。为解决数据重构时存在的时序信息损失问题,利用重排自回归保留序列的时序信息。然后,对于重构后的条件片,根据相邻条件片协整模型的相似度合并条件片,捕捉条件轴上协整关系的变化,将协整关系相似的数据划分至同一工况,并建立统一的协整模型。最后,利用某风电场数据对所提算法的有效性进行了验证,并通过与传统线性协整算法的比较说明其优越性。

Aiming at the problem of nonstationary feature extraction caused by the nonstationary variables obeying the nonlinear cointegration relationship in the industrial production process,a condition-driven nonlinear cointegration algorithm is proposed.First,according to the operating state of the device,the data is reconstructed from the time axis to the condition axis,so that the data in each condition slice have similar process characteristics.In order to solve the problem of timing information loss during data reconstruction,rearrangement autoregression is used to retain the timing information of the sequence.Then,for the reconstructed condition slices,the condition slices are merged according to the similarity of the cointegration model of adjacent condition slices,the change of the cointegration relationship on the condition axis is captured,and the data with similar cointegration relationship are divided into the same working condition and a unified cointegration model is established for it.Finally,the validity of the proposed algorithm is verified by the data of a wind farm,and its superiority is demonstrated by comparison with the traditional linear cointegration algorithm.