To deal with the high-dimensional source data，a new method for identifying the coherent generators is proposed based on the non-negative matrix factorization（NMF）algorithm.Taking the angular speed of the generator as source data，its dimension can be reduced by using NMF algorithm.Because of the non-negative property of the obtained lower-dimensional matrix，it possesses some prior properties such as eliminating redundant data，reducing dimension，etc.，and could keep the practical significance of the original problem.The normalized lower-dimensional matrix is then clustered by K-means method to achieve grouping of coherent generators.Finally，simulation results are performed on the New England10-machine39-node system，some comparison results about grouping performance between NMF algorithm and principal component analysis（PCA）algorithm are obtained，which could show the effectiveness of the proposed method in the identification of coherent generators.