【作者】 王俊杰；

【导师】 郑益慧；

【作者基本信息】 上海交通大学 ， 电工理论与新技术， 2011， 硕士

【摘要】 随着用电设备对电能质量要求的提高,谐波抑制和无功补偿问题越来越受到广泛的研究和重视,而以谐波抑制、无功补偿为主要功能的有源电力滤波器(Active Power Filter, APF)的基本理论已经成熟,其采用现代电力电子技术、数字信号处理技术和先进控制理论,被视为现在解决谐波污染的最有效和最具潜力的途径。本文在综合国内外相关文献基础上,介绍了有源电力滤波器的发展历程、现状和趋势,以及有源电力滤波器的分类和拓扑结构,对单相并联型有源电力滤波器的谐波检测和并联型有源电力滤波器的控制策略两个主要部分进行了研究,仿真和实验验证了方法的有效性和合理性。首先,介绍了有源电力滤波器的发展及研究现状,并对谐波产生的原因及其危害和谐波检测算法做了概括性的论述和总结。其次,通过对并联型有源电力滤波器的工作原理和工作特性的分析,提出了三种谐波及无功电流检测方法。鉴于传统自适应谐波检测算法中步长的选择需要在收敛速度与稳态精度之间进行折衷考虑,降低了算法的实用性。为了克服此局限性,首先,本文提出动态因子LMS (Dynamic Factor Least Mean Square, DFLMS)自适应谐波检测算法,通过引入动量项,并利用误差信号在相邻时刻的时间均值估计来控制步长更新,大大加快了算法的收敛速度。采用动态因子项对误差进行再次调解,减小了低信噪比情况下的稳态失调,增强了算法对噪声的抗干扰性。其次,针对DFLMS算法的基本原理和理论推导做了详细的阐述,并通过仿真和实验将DFLMS算法与其他自适应谐波检测方法在有源电力滤波器中的应用进行了对比。在此基础上对算法进行改进,提出多重因子变步长(Multi-Factor Least Mean Square, MFLMS)自适应谐波检测算法,该方法增加了相邻时刻误差信号的自相关值及静态因子对权值迭代的影响,并利用误差信号的时间均值估计来控制步长更新,加快了谐波检测的动态响应速度。最后,为了进一步提高跟踪后的精度,在一定的时间后增加相邻时刻误差信号的自相关值及静态因子对权值迭代的影响,提出了限幅变步长LMS(Limiting Least Mean Square, LLMS)谐波检测算法,对算法进行了进一步的改进。结果表明上述三种自适应谐波和无功电流检测方法具有实现简单、响应速度更快、稳态精度更高的优点。最后,对有源电力滤波器进行了应用研究。在对有源电力滤波器的主电路进行研究的基础上,给出功率器件的选取原则和主要电路参数的计算方法,基于谐波电流检测算法和电流跟踪控制策略作了软件的实现,为有源电力滤波器的实际应用提供了有益的探索。更多还原

【Abstract】 With the electrical equipments’high requirements for power quality, harmonic suppression and reactive power compensation are increasingly paid attention, and the basic theory of active power filter (Active Power Filter, APF) is matured, which has the function of harmonic suppression and reactive power compensation. It adopts modern power electronic technology, digital signal processing technology and advanced control theory, which is considered as the most effective and most promising ways to solve the harmonic pollution. Based on the comprehensive basis of relevant literature, it describes the course of development, current status and trends of active power filter in this thesis. It also introduces the classification of the active power filter’s topology. In the thesis, it proposes the harmonic detection of single-phase parallel active power filter and control strategy as two main parts of research. In the end, simulation and experiment verify the validity and rationality of the methods above.Firstly, it introduces the development and research status of the active power filter, and gives a general discussion and conclusion on harmonic causes and the hazardous of the harmonic.Secondly, through the working principle of active power filter and characteristics of the analysis, three kinds of harmonic and reactive current detection methods are presented. For the traditional adaptive harmonic detection algorithm, it has to select the parameters between the desired step size and speed of convergence trade-off, which reduces the practicality of the algorithm. To overcome this problem, a novel Dynamic Factor Least Mean Square (DFLMS) algorithm is proposed for APF (Active Power Filter)’s harmonic current detection, which is based on adaptive noise cancellation technology and takes the single phase APF as a platform. With the consideration of poor performance of the traditional algorithm, the momentum term is introduced, and the error signal’s mean estimate is used in the adjacent time point to control the updating of the step size, which greatly accelerates the convergence speed. A dynamic factor is introduced to not only accelerate the dynamic response speed, but also for re-mediation of the error term. Then it gives a detailed description of the basic principle and theoretical analysis of the DFLMS algorithm. Through the simulation and experiment, DFLMS algorithm and other adaptive harmonic detection methods were compared in active power filter application. On this basis, to improve the algorithm’s performance, the multiple factor least mean square variable step size (Multi-Factor Least Mean Square, MFLMS) adaptive harmonic detection method is proposed. It increases the error signal adjacent shike static self-related factors on the right values and the impact of value iteration, which uses estimates of the error signal to control the time step update. It not only speeds up the dynamic response speed, but also reduces the steady-state offset in low SNR conditions. Finally, to improve the tracking accuracy, after a certain time, the error signal adjacent shike autocorrelation value and the value of the static weight of the iteration are increased, it proposes a limiting least mean square LMS (Limiting Least Mean Square, LLMS) variable step size adaptive harmonic detection algorithm. The results show that the three adaptive harmonic and reactive current detection methods are simpler, faster response and better steady state accuracy.Finally, the application research of the active power filter is realized. Based on the research of the main circuit of active power filter, it gives the selection principle of the power devices and the calculating of main circuit parameters. It makes the realization of the software based on the harmonic detection and current tracking control strategy, which provides a useful exploration for the practical application of active power filter.更多还原