The clustering is one of important methods in multivariate analysis. The clustering of time series is also important and several clustering methods are available. Recently, a k-means type method was proposed for trends of time series. In this method each object for clustering consists of univariate time series. In this study, a generalization of this method is proposed for the case where each object consists of multivariate time series. The applicability of the proposed method is examined by simulation studies. Moreover the clustering of time series on COVID-19 cases is considered by applying the proposed method.