Harmonic Detection Based On Improved Independent Component Analysis Algorithm

In recent years, with the rapid development of economy and technology, due to the increasing power system load and the extensive use of nonlinear devices, making the power system harmonic pollution is getting worse. Harmonic pollution produced by the interference of the power grid is not only a serious power quality, but also affects the quality of people's daily electricity. In order to effectively control power system harmonics, harmonic testing needs to determine the harmonic components contained in the grid.Independent component analysis as the rise of these recent years, an efficient blind source separation method has been applied to image processing, speech processing, seismic signal processing, etc., but in the power system harmonic detection applications is not very common. This thesis will introduce the independent component analysis to harmonic detection; so as to power system harmonic detection provides a new idea. The main contents are as follows:1. This thesis introduces the research background of independent component analysis and the domestic and foreign development situation. Given the concept of independent component analysis and some basic knowledge which solving the process. On the basis of based on negative entropy is discussed based on kurtosis FastICA algorithm and the FastICA algorithm and the two algorithm respective characteristics.2. This thesis introduces the FastICA algorithm based on negative entropy, and the algorithm of the original basis, the algorithm was improved, and proposes an improved algorithm based on negative entropy of FastICA and applied to the harmonic detection. The improved algorithm to Newton iterative method was improved, making the improved algorithm satisfy third-order convergence and accelerate the convergence speed. Simulation experiments indicate that the improved based on negative entropy of FastICA algorithm is obviously superior to the FastICA algorithm based on negative entropy.3. This thesis introduces the FastICA algorithm based on kurtosis, and the algorithm of the original basis, the algorithm is proposed based on kurtosis dynamic ICA algorithm and applied to the harmonic detection. This algorithm overcomes the original based on kurtosis FastICA algorithm in each iteration process are required to use the whole sample data shortcomings, will kurtosis iterative equation with the original based on kurtosis of FastICA algorithm combining, making the improved algorithm in each iteration process of each iteration simply use single sample data can, improved algorithm of real-time. Through the simulation result shows based on kurtosis dynamic ICA algorithm based on kurtosis is obviously superior to the FastICA algorithm.