Mixed Over-Voltage Decomposition Using Atomic Decompositions Based on a Damped Sinusoids Atom Dictionary

The main purpose of this paper is to establish a signal decomposition system aiming at mixed over-voltages in power systems. In an electric power system, over-voltage presents a great threat for the system safety. Analysis and identification of over-voltages is helpful to improve the stability and safety of power systems. Through statistical analysis of a collection of field over-voltage records, it was found that a kind of complicated signals created by mixing of multiple different over-voltages is difficult to identify correctly with current classification algorithms. In order to improve the classification and identification accuracy of over-voltages, a mixed over-voltage decomposition system based on the atomic decomposition and a damped sinusoid atom dictionary has been established. This decomposition system is optimized by using particle swarm optimization and the fast Fourier transform. Aiming at possible fault decomposition results during decomposition of the over-voltage signal, a double-atom decomposition algorithm is proposed in this paper. By taking three typical mixed over-voltages as examples, the validity of the algorithm is demonstrated.