Classification of power-quality disturbances using PSO-MP and parametric dictionaries

This paper aims to develop a new scheme for the classification of power-quality disturbances (PQDs). We propose to employ two discriminative dictionaries, designed based on the structures of PQDs, to respectively decompose a disturbance signal. Matching pursuit optimized by particle swarm optimization (PSO-MP) is used as the decomposition method. Reconstruction errors after sparse coding are employed to coarsely classify the PQDs into two categories, corresponding to the two dictionaries. Next, the specific class can be identified by evaluating the value of parameters of atoms. One main advantage of the approach is that it does not require a training set as many other classification methods do. The PQDs considered in this paper include sag, swell, interruption, harmonic and oscillatory transient. Experimental results indicate that the proposed approach achieves a high classification accuracy and robustness against noise.