【作者】 刘传林；

【导师】 刘开培；

【作者基本信息】 武汉大学 ， 电力电子与电力传动， 2012， 博士

【摘要】 应用有源电力滤波器(Active Power Filter, APF)提高电能质量,需要获取电网相关电流分量如畸变电流或谐波,实时精确的测量有利于提高有源电力滤波器的性能。实时精确的电网畸变电流或谐波检测是电能质量控制的重要问题。本文选择LMS (Least Mean Square)自适应算法作为电网畸变电流检测系统的核心环节进行理论与算法研究。在理论方面,阐明了基于LMS算法自适应系统的线性时变特性,分析了电网畸变电流检测系统的瞬态特性；由于迭代步长是自适应算法的关键因素之一,本文研究了自适应LMS算法的步长选择原理。通过分析基于LMS的自适应电网畸变电流检测算法的缺陷,提出了改进型LMS算法并研究其在电网畸变电流检测中的应用,这些算法包括：1)基于嵌入低通滤波器的改进型自适应畸变电流检测算法,运用低通滤波器滤去权值或误差迭代中的高频分量,提高算法稳态精度与收敛速度。分析嵌入均值滤波的aLPF-LMS算法自适应谐波检测的稳定性。2)研究了OVS-LMS与ROVS-LMS变步长算法,提高检测的稳态精度,改善了收敛速度。3)基于双LMS构成自适应组合系统,通过兼顾自适应的快速性能与稳定性能改善电网畸变电流的检测。4)通过在自适应系统输入信号环节预置带通滤波器,改善LMS算法电网畸变电流检测的稳态误差,提高检测收敛速度。研究改进LMS算法在单相谐波检测中的应用。基于LMS算法探索三相系统畸变电流的自适应检测方法,研究了三相平衡系统与三相不平衡系统的自适应畸变电流检测问题。研究了基于瞬时无功功率理论的自适应三相电网谐波检测方法,包括改进型LMS算法自适应三相谐波检测方法与基于重采样原理的自适应三相谐波检测方法,克服了基于瞬时无功功率理论实现三相谐波检测的缺陷,提高了三相谐波检测的稳态精度与跟随性能。本文对提出的基于LMS改进算法的单相以及三相自适应畸变电流与谐波检测方法进行了仿真验证,分析基于LMS改进算法自适应畸变电流或谐波检测的稳态精度与动态跟随性能。最后,对改进的LPFw-LMS、LPFd-LMS、LPFwd-LMS算法进行了实验分析,测试了改进型自适应畸变电流检测算法的性能。更多还原

【Abstract】 In APF (Active Power Filter), it is necessary to get distorted current components or harmonics in electric distribution network for suppressing distorted currents or harmonics, and real-time and accurate detection of currents can improve APF’s quality. So real-time and accurate distorted current detection is an important problem in electric power quality control.This dissertation made a selection which employs least mean square (LMS) adaptive algorithm to distorted current detection system as the key, and discussed theories and algorithms of the adaptive system. In theory, research included some characters of linearity and time-varying of the adaptive system based on LMS algorithm, and instantaneous characters of adaptive distorted current detection system. For the key facter of adaptive algorithm, this dissertation proved principle of selecting step-size to accelerate convergence based on LMS algorithm. This dissertation discussed errors and drawbacks of adaptive distorted current detection system based on LMS, and proposed methods to improve performance of distorted current detection in power system. These methods included:1) Improved adaptive distorted current detection method based on embedded low pass filter. By filtering fluctuation of weights and errors in LMS algorithm, the adaptive system can detect distorted currents timely and accurately. This dissertation analyzed stability of an LMS algorithm which adopts an average filter to eliminate fluctuation (aLPF-LMS) in harmonic detection.2) Optimal variable step-size LMS algorithm (OVS-LMS) and recursion optimal variable step-size LMS algorithm (ROVS-LMS).The algorithm can improve adaptive convergence speed reduce errors of steady state in distorted current and harmonic detection.3) Combined LMS which improves precision of distorted current detection. When keeping dynamics of detecting distorted current, the combination can reduce errors of steady state with twin-LMS.4) Employing band pass filter (BPF) to pretreate input of adaptive system based on LMS algorithm. The LMS algorithm can reduce the errors of steady state of harmonic current detection and improve convergence speed of system. This dissertation also discussed adaptive harmonic detection methods based on improved LMS algorithm.This dissertation discussed adaptive distorted current detection methods of3-phase circuits based on LMS algorithm, and discussed distorted current detection of3-phase balance system and3-phase unbalance system. For three phase symmetrical system, this dissertation discussed harmonic detection methods based on the instantaneous reactive power theory and adaptive LMS algorithm. These methods include adaptive3-phase harmonic detection method based on improved LMS algorithm and resampling LMS algorithm, overcoming the drawback of harmonic detection based on the instantaneous reactive power theory, improving dynamic performance and reducing the steady state errors in3-phase harmonic detection.In these studies, simulations of adaptive distorted current detection or harmonic detection methods based on modified LMS algorithms have been done by using MATLAB, dynamics and steady state errors of these algorithms have been analyzed, including3-phase circuits.Finally, modified algorithms including LPFw-LMS, LPFd-LMS and LPFwd-LMS have been tested and verified by experiments. According to experimental results, these algorithms are proved to detect distorted current efficiently.更多还原