It is difficult to identify the categories of power quality hybrid disturbances (PQHDs) because of the complex characteristics and the feature overlap of PQHDs, for which reason a novel proposal provided in this correspondence is to identify the PQHDs fast. S transform and TT transform are applied to extract the 60 feature quantities of 15 classes of PQHD signals produced by mathematic models. Principal components of feature set are acquired by principal components analysis(PCA). The one-versus-rest support vector machine is constructed to identify the kinds of PQHDs by introducing PSO which is used to optimize the penalty factor and slack variable. Ultimately, the disturbance signal data is produced and the PSO-SVMs classifier is established based on MATLAB, the results of simulation verify that the proposed approach is reliable and stable.