【作者】 张海霞；

【导师】 谭阳红；

【作者基本信息】 湖南大学 ， 电气工程， 2019， 博士

【摘要】 大力发展新能源是实现可持续发展的重要途径。为获取丰富的风资源,更多的风电场被建设在沙漠、海岛或者高山峻岭等较恶劣的环境之中。以永磁同步风力发电机(Permanent Magnet Synchronous Generator,PMSG)为代表的全功率并网系统是目前主流的风电并网拓扑之一。其主要部件全功率变流器内含多个电力电子器件,容易受环境影响发生故障。据统计变流器故障造成的经济损失已达总成本的13%。变流器故障中,开路故障是最常见故障之一,且不易发现。PMSG系统变流器一旦发生开路故障,可能会使PMSG系统长期处于非正常运行状态,给其他电力电子器件施加工作压力,埋下二次故障的隐患,甚至威胁到整个电力系统的安全。因此,剖析变流器开路故障后的系统动态特性,精确定位机组变流器故障是确保系统安全运行的重要工作。本文对现今变流器开路故障动态特性研究分析理论进行了扩展,并以数据分析为基础实现了PMSG系统变流器开路故障诊断,具体工作如下。电流断流被忽略,多管开路故障分析较少等问题,使以往PMSG系统变流器开路故障下的电压动态特性研究不够完善。在充分考虑断流情况和多管开路的前提下,以开关函数PMSG系统模型为基础,较全面地揭示了PMSG系统变流器开路故障下电压动态特性。在研究过程中,以单管开路及非同相多管开路两大类故障模式为例进行深入分析,充实了多管开路故障的研究。并以故障相电流极性及其通断情况为依据,将单管和多管开路分别分成3种和8种运行方式进行,完善了电流断流情况下的系统特性分析,推导了下管电压、直流侧对地电压和定子电压的数学表达式,较全面地揭示了系统变流器开路故障下的动态特性。通过实验仿真,验证了方法的准确性和有效性。针对因风电系统的强非线性性和耦合性,故障后三相电流难以量化分析的问题,提出了一种基于混合逻辑动态模型(Mixed Logic Dynamic Model,MLD)的电流特性分析方法。在PMSG系统的动态逻辑关系分析基础上,建立了包括同相多管开路在内的多种故障下MLD模型,弥补了开关函数模型在同相双管开路运行特性分析中的不足。将通常被忽略的触发脉冲动态变化分析纳入研究,获得了各脉冲组合概率变化规律。通过PMSG系统变流器闭环控制策略及PWM控制策略分析,解决了以往电流分析中因系统强耦合性难以执行的问题。通过算例仿真与测试,验证了方法的准确性和有效性。基于数据分析的故障诊断方法对外部特征十分敏感,但以往的研究鲜少对信号自身特征进行深入分析,外部特征突显困难,无法确保故障诊断率。针对这一问题,提出了一种计及外部故障特征突显精准捕捉的变流器开路故障诊断方法。该方法立足于信号自身特性的研究,对比分析了信号负载特性、抗干扰能力、特征分离容易程度等信息,为故障特征提取信号选取提供了依据;深入研究了所选信号的外部特征突显时间、突显条件以及突显频率,确保了外部特征的捕捉,解决了外部特征突显困难的问题。基于信号特性的研究结果,提出了一种基于下管电压局部有效值的故障诊断方法,避免了阈值的选取和电压偏差的求取。测试与分析结果表明利用下管电压局部有效值的故障诊断方法可以对不同情况下的单管和多管开路故障进行正确诊断。电流故障诊断法无需增加传感器,相比电压故障诊断法有着其独特的优势,但因电流信号对负载及外界环境其他因素敏感,致使电流故障诊断方法效果常低于电压故障诊断法。为增加电流特征的抗干扰能力,从三相电流相互间的约束关系出发,结合神经网络在非线性系统中的优势,提出了一种基于三相电流相关性的神经网络故障诊断方法。该方法以三相电流相关性特征作为神经网络的输入,减小了电流负载特性对诊断结果的影响,计算量较少,且所需的存储空间不大。对比基于DQ电流和三相电流波形的神经网络故障诊断方法,基于三相电流相关性的神经网络故障诊断方法在诊断率、抗干扰能力、对神经网络结构及算法的依赖性等方面均有明显的优势,具有一定工程应用价值。更多还原

【Abstract】 Developing new energy sources is an important way to meet sustainable development.Wind farms are usually in deserts,islands,and mountains.These surroundings are complex and bad for the safe of wind farms.The full power gridconnected system with a permanent magnet synchronous generator(PMSG)is one of the mainstream wind power grid-connected topologies.The full power converter,as one of the main components of PMSG system,has many power electronic devices,which hurt easily in the harsh environment.According to statistics,the economic losses caused by converter faults have reached 13% of the total cost.In converter faults,the open-circuit fault(OCF)is one of the most common faults and are not easily found.If an OCF occurs and exists secretly,the other power devices will get much pressure,the risk of secondary faults rise,and the security of the power grid is threatened.Therefore,analyzing the dynamic characteristics of the system after an OCF in a converter,and locating the fault accurately are much important tasks for the safe of the wind power system.On the basis of the existing research and analysis theory of dynamic characteristics for an OCF in a converter,this dissertation reveals the dynamic distortion characteristics and studys on the fault diagnosis for the faults in a converter of PMSG system.In the previous research on the voltage characteristics of the PMSG system after an OCF happened in a converter,the current in the zero-crossing section is often ignored,and the multi-switch OCFs are not analyzed.For remedying these problems,both of the single-switch faults and the multi-switch faults are taken as examples.The study of the multi-switch OCF is enriched.Besides,the research for all the examples are classified by the polarity of the faulty phase current and its on-off condition.So,the single-switch OCF and the multi-switch OCF are analyzed in 3 and 8 kinds of operation modes,respectively.The characteristic analysis during the zero-crossing section of stator current are improved.So,the operating characteristics of PMSG system after an OCF happened in a converter are revealed,and the mathematical expressions of the lower tube voltage,the DC link to ground voltage and the stator voltage are obtained.The accuracy and effectiveness of the method are verified by experimental simulations.Due to the strong nonlinearity and the coupling of the wind power system,the analysis for three-phase current is difficult.Aiming at this problem,this paper proposes a current characteristic analysis method based on Mixed Logic Dynamic Model(MLD).After the dynamic logic relationship analysis of PMSG system,several MLD models under different fault modes,surely,containing under a one-phase multi-switch OCF mode,are obtained.That makes up for the shortcomings of the switch function mode l in the analysis of the one-phase multi-switch OCFs.The dynamic analysis of the trigger pulse that is usually ignored also is included in this study.And the rule of probability variation for each pulse combination is obtained.In order to deal with the coupling relationship between the system and the external environment,the closed loop control strategy for the converter in PMSG system and the PWM control strategy are analyzed,which solves the problem that the current analysis is difficult to do.The data-based fault diagnosis method is always sensitive to external features.However,in the past research,the characteristics of signal are rarely analyzed in-depth,which leads to difficulty highlight the external features.So the data-based fault diagnosis method cannot be ensured the high perform.Due to this reason,a fault diagnosis method is proposed after the signal analysis in-depth.In the process of signal analysis,many features,such as the load characteristics,the robustness,the separation,are compared and analyzed during an OCF occurred in PMSG system.Then,the appearance time,appearance condition and appearance frequency of external feature in signal are studied to make ensure the capture of external features in-depth.According to the research results of signal,a voltage-based fault diagnosis method is proposed.This method does not need threshold.The calculation of voltage deviation is unnecessary.The test and analysis results show that the proposed voltage-base method can correctly detect the single-switch and multi-switch OCFs under different wind speed.the current-based method does not require additional sensors.However,the external feature also affects the perform of a current-based fault diagnosis method.The current is sensitive to the load and other factors from environments.The fault diagnosis ratio of a current-based method always shows lower than the voltage-based method.So,in order to enhance the robustness of the current signal,this dissertation focuses on the constraint relationship between three-phase currents.A novel current-based neural network fault diagnosis method is proposed.This proposed method takes the three-phase current correlation features as the input of the neural network.Therefore,the diagnosis result shows with little effects from the current load,and its achievement does not need a large calculated amount.The required storage space also is not large.Compared with the neural network fault diagnosis method based on DQ current and three-phase current waveform,the proposed method has obvious advantages in diagnosis rate,robustness,and dependence on neural network structure and algorithm.更多还原