【作者】 裴喜平；

【导师】 郝晓弘；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2014， 博士

【摘要】 随着现代电能质量问题的日益严重,用户对电力系统的电压供电质量提出了越来越高的要求,能够抑制电力系统中各种电压扰动以保证用户供电质量的补偿装置得到了快速发展。在各种现代电能质量问题中,电压跌落是目前最为突出、也是最为严重的动态电压质量问题,动态电压恢复器(DVR)作为最经济、有效的治理电压跌落的定制电力电子装置,在电能质量的控制技术领域成为研究的热点。本文以解决DVR中相关的检测、控制技术为主线,首先研究了单相幅相锁相环(1MPLL)的设计方法,其主要应用于单相DVR或三相四线制配电网需要以三相独立方式补偿的DVR设计中；然后研究了三相幅相锁相环(3MPLL)的设计方法,主要应用干三相三线制配电网DVR的设计中：在利用上述幅相锁相环解决了电压跌落识别和基于相位的参考补偿量生成后,研究了动态电压恢复器的控制策略；最后对在电能质量监控(包括电压跌落)和评估中广泛使用的三相电网对称分量估计进行了研究。基于目标函数优化的思想提出了一种单相幅相锁相环的优化设计方法。该算法通过梯度下降法分析了目标函数的最速下降方向,确保了锁相环系统的全局快速收敛,通过目标函数的恰当选择和系统参数的合理设计,幅相锁相环能在各种干扰下确保对基频信号特征量的快速精确跟踪,如幅值、相位和频率等,具有良好的频率自适应性以及抗谐波干扰能力。本文所提单相锁相环结构简单,物理意义明确,参数易于设计,在提供配电网网侧电压同步信息的同时,也能提供对电压跌落幅值的检测,因此特别适合应用于DVR系统中,DVR系统利用检测到的幅值信息决策是否启动补偿,在确定电压跌落的情况下根据锁相环提供的相位信息利用某种补偿算法实时生成参考补偿输出。传统DVR系统中一般采用幅值和相位各自独立的检测算法,势必增加了系统的计算量,不利于DVR系统的在线计算,本算法更有利于实际实施。基于模型参考自适应思想提出了一种三相幅相锁相环设计方法。算法设计了具有频率自适应的三相三线制配电网正、负序电压分量状态观测器,利用李雅普诺夫稳定性理论进行自适应率设计,满足了锁相环正、负序分量估计全局渐近稳定的要求,理论上实现了正序分量幅值和相位的检测不受负序分量的影响,且解决了SFR-PLL分析中未对全局稳定性讨论的不足,系统结构简单,有利于系统关键参数的设计。算法由于事先考虑到频率波动对锁相环精度的影响,因此,系统对频率波动有很强的鲁棒性。讨论了DVR主电路中逆变电路、能量提取电路、滤波电路中LC参数的设计准则以及电容储能条件下延长跌落补偿时间的优化补偿策略,在此基础上主要针对目前DVR控制方法存在鲁棒性不足的问题,提出了基于模型预测控制(MPC)和鲁棒H控制的控制器设计方法。基于非参数预测模型(脉冲响应或阶跃响应)的预测控制,算法复杂,在线运算量大,需整定的参数较多,难以直接应用到快速的实时控制系统领域。本文采用双环控制器结构,首先通过内环电流控制器对DVR对象进行改造,在增广控制对象的基础上获得了简化的预测模型,基于此进行模型预测控制器的设计。模型预测控制通过滚动优化、反馈校正等措施降低了模型误差对性能的影响,控制器根据系统未来的预测输出状态确定当前的控制动作,因此可以减小超调,提高系统的响应速度,对输出偏差的滚动优化,确保了系统具有良好的稳态精度,满足了DVR快速、准确补偿电压跌落的目的。鲁棒H∞控制由于事先充分考虑了被控对象的参数不确定因素,因此控制系统具有很强的鲁棒性能。本文针对DVR主电路中LC滤波器参数摄动对被控对象模型造成的不确定性,在考虑鲁棒稳定性、鲁棒跟踪性能的基础上,根据H∞控制理论设计了控制器,通过恰当的选取界函数,使DVR系统在面对参数摄动时具有很强的鲁棒性能。基于最小均方学习规则神经网络提出了一种三相电网对称分量的实时估计方法,检测获得的对称分量可以应用于电压跌落识别以及其他电能质量监控中。建立了三相电压的对称分量模型,对最小均方学习规则神经网络根据三相系统对称分量估计的需要进行了合理改进,通过神经网络的学习过程完成了对各对称分量的实时估计。算法具有计算量小,动态响应速度快和精度高的特点,能满足电力系统实时计算的需要。更多还原

【Abstract】 With the modern power quality problems growing, more and more users have higher requirements for the voltage quality of power system, the compensation devices has been developed rapidly to suppress a variety of voltage disturbances in power system. Among the modern power quality problems, the voltage sag is one of the most prominent and serious dynamic voltage quality problems. Dynamic Voltage Restorer (DVR) as the most economical and effective custom power electronic devices has become a research hotspot in the power quality control technology. The main purpose of this papar is to research the detection and control technologies in DVR. At first, we study the design approach of the single phase amplitude-phase locked loop(1MPLL), which is mainly used in the single-phase DVR or the three-phase independent compensation DVR that is applied to three-phase four-wire power distribution network; Then, A design approach of the three phase amplitude-phase locked loop(3MPLL) is proposed to design the three-phase three-wire DVR; Furthermore, After identifying the voltage sag and generating the reference compensation with the above amplitude-phase locked loop, the control strategy about the DVR are explored; Finally, we estimate and analysize the symmetrical components which are widely used in power quality monitoring and evaluation system.An optimized design approach about1MPLL is presented in this paper based on the ideas of objective function optimization; it is used to analyze the steepest descent direction of the objective function with the gradient descent algorithm, which ensures a fast and global convergence of phase-locked loop system. With reasonable objective function and appropriate parameters, amplitude-phase locked loop can fast and accurately track on the features (such as amplitude, phase and frequency and so on) of the fundamental frequency signal under interferences, which has the better abilities in frequency adaptability and anti-harmonic interference. Single phase locked loop presented in this papar has the advantages of simple structure, clear physical meaning and convenient parameter design, and provides synchronization information of distribution network voltage as well as the amplitude of voltage sag. Therefore, it is suitable for DVR system particularly. The DVR decide whether to initiate compensation with the detected amplitude information, when the voltage sag is appeared, the compensation algorithm is applied to generate real-time reference compensation output signal based on phase information from phase-locked loop. Compared with the traditional DVRs which detect amplitude and phase separately and inevitably increase the amount of computation, can not be used in real time, the proposed algorithm is more conducive to the actual implementation.A design method of a three-phase locked loop based on the conception of adaptive model reference is proposed, we designs positive and negative frequency adaptative sequence voltage component state observer and frequency adaptive rate by Lyapunov Stability Theory. This method provides the global asymptotic stability to the phase-locked loop and can detect the amplitude and phase of the positive sequence component; it is not affected by the negative sequence component theoretically, and avoids the disadvantages of SFR-PLL, in which the global stability is not discussed. The system structure is simple and conducive to set the system parameters, the algorithm takes into account the impact of frequency fluctuations on the phase-locked loop accuracy. Therefore, the system has a strong robustness for frequency fluctuations.The controller design method based on model predictive control and robust Hn control theory is presented to offset the insufficient robustness of the current DVR control methods. Predictive control methods based on non-parametric (impulse response or step response) models are complex with large amount of computation and more tuning parameters, it is difficult to be applied in the real-time control systems. In this paper, the DVR is altered by an inner current controller, and a predictive controller is designed with a simplified prediction model on the augmented control object. The model predictive control method reduces the impact of model errors on the performance of scroll optimization, feedback correction and other measures. The controller determines the current control action based on the future output state of the system; therefore, it reduces overshoot, improves the response speed of the system and meets the fast and accurate voltage sag compensation purposes of DVR. Since robust Hx control considers the uncertainty of controlled object parameter in advance, the control system has a strong robust performance. The uncertainty of the controlled object model is casued by LC filter parameter perturbations in DVR main circuit, we select the appropriate boundary function by considering the robust stability and robust tracking performance and designs the controller based on Hx control theory. The controller makes the DVR system has strong robustness against the parameter perturbationsA real-time symmetrical component estimation method of three-phase system based on minimum mean square learning rule neural network is presented; the symmetrical components can be used to identify voltage sags and other power quality monitoring applications. A symmetrical component model of three-phase voltage is established, the minimum mean square learning rule neural network is revised according to symmetrical component estimation of three-phase system, the algorithm estimate the symmetrical components by the neural network in real-time, which has less calculation, faster dynamic response and higher accuracy, can be used in power system in real time.更多还原