Information process and its application of power system relate widely to the technology of data mind and character extraction. In order to increase the efficiency of clustering algorithm, a novel clustering algorithm of power system, which combines particle swarm optimization (PSO) with fuzzy C means (FCM) organically, is presented. The algorithm uses PSO to replace the iteration process of FCM based on the gradient descent. So it has strong global searching capacity, and avoids the problems of local optimization and initialization values. At the same time, the kernel method is involved to map the eigenspace of low dimensions to that of high dimensions by kernel function, so the optimization of eigenvalue is enhanced, and the eigenvalue is easier to be clustered in the eigenspace of high dimension. The simulation result of power system load clustering shows that the proposed algorithm is more accurate and efficient than FCM.