Adaptive Harmonic Detection Method In Power System

Harmonic pollution problem becomes a long-term problem in power system. As with progress of technology, the power system is full of a variety of non-linear loads in most cases. And the harmonics pollution can deteriorate owing to the new concepts such as smart micro-grid. Thus, the power quality analysis has to consider the impact of power system harmonics. By using the modern power electronics technology, digital signal processing technology and advanced control theory, Active Power Filter (APF) as a widely used method suppresses the harmonic pollution. The main functions of APF include harmonic suppression and reactive power compensation. However, the performance of APF largely relies on fast and accurate detection of the harmonic current components.To solve these problems, the following work has been done in this dissertation.Firstly, this dissertation briefly introduces some concepts of harmonic, the impact of harmonics on power system and some control measures. Moreover, the dissertation indicates that the harmonic current detection is important to analysis harmonic, and summarizes the present situation of harmonic current detection method.Secondly, the dissertation describes three common adaptive algorithms, namely, least mean square(LMS) algorithm, recursive least square(RLS) algorithm and kalman filtering(KF) algorithm, and provides a simple analysis of their features. Besides, it elaborated the principle and structure of the adaptive harmonic current detection method which is based on adaptive noise cancellation theory. All the adaptive algorithms are used in the method, comparison with each other by the steady-state error, convergence speed and computing complexity.Then, this dissertation analyzes and summarizes the advantages and shortcomings of the harmonic current detection method of traditional LMS algorithm and variable step size LMS algorithm. A new variable step size LMS algorithm for harmonic current detection method is proposed. This method uses the time coherent average to weak the influence of harmonic current on the step size updating, and based on a Witch of Agnesi function change the parameters of the momentum in the weight equation. During the initial period or mutation of the system, the momentum has a greater impact on the weight updated formula. When the system reaches a steady state, the parameter of momentum converge to a small number, to reduce the momentum’s impact on the weight updated formula. This approach not only improves the convergence speed, but also ensures the steady-state error, and the algorithm achieves a good tracking performance.Finally, this dissertation presents a harmonic current detection method based on affine projection algorithm (APA). As compared with LMS, it is observed that both algorithms have a substantially identical the steady-state error, but the APA has a shorter dynamic response time. Additionally, the effect of the affine projection order is shown on dissertation. Then, a variable step size is used in the APA harmonic current detection algorithm, which significantly improves the convergence rate than the original algorithm and its steady-state error essentially unchanged.Simulation results using MATLAB software to verify the performances of proposed algorithms in this dissertation.