Study On Separation Of Harmonic Sources Based On Independent Component Analysis

Recent years, with the development of power electronics technology and the mass usage of non-linear loads, more and more harmonic are injected into the network。It corrupts the power quality of the electric network badly. In order to control the harmonic effectively and reduce the harmonic pollution, we need to have an evaluation of the distribution of harmonic sources as well as the transmission state and the harm degree of harmonic.Independent component analysis (ICA) is a multi-channel signal processing method developed from blind source separation technology in the late 1990s. It can estimate the wave of source signals under the condition of having no prior information of source signals. It has been widely used in biomedical signal processing, image processing, speech processing, and other aspects, but its application in the power system has only just begun. The thesis adopts ICA to separate harmonic sources when unknowing the topology and parameters of network. It provides a new idea for harmonic source separation with incomplete network information. The main work and conclusions are as follows:①It summarizes the correlative concepts of harmonic, and analyses the advantage and disadvantage of the existing methods of harmonic source separation. After a thorough study of these methods, we find that when the network information is incomplete, the existing methods cannot meet the requirements. So we need to introduce new research ideas to break the current bottle-neck.②It introduces the theory and application of blind source separation and researches the theory and mathematical models of ICA. The estimated principle of ICA is focused on.③The calculation model of harmonic source separation is established. It finds out the consistency of harmonic source separation model with the ICA model. And a new separation method of harmonic sources which is based on ICA is proposed. This method makes use of the statistical independence of harmonic sources to separate the harmonic source under the condition of unknowing the parameters and topology of network. Through the analysis of Infomax algorithm and FastICA algorithm and comparison between them, the thesis selects FastICA algorithm to separate harmonic sources, and gives the step of harmonic separation. ④On the basis of theory analysis, the simulation is carried out on the IEEE-14 bus harmonic test system, and the performance of simulation result is evaluated. Corresponding example and performance index indicate: it can effectively separate the harmonic source by using ICA. Further more, whether the harmonic collections contain the harmonic source or not, simulation has the same effect. That is, the selection of measurement location has no influence on simulation result. In addition, the stronger non-Gaussian of the harmonic source, the better effect of separation.