光伏系统直流电弧故障检测及保护策略研究

With the constant growth of PV installed capacity, fires arising from DC arc fault in PV system are occurring more often. Closely related to the problem of arcing is the fact that the DC-part of PV-installations can not be powered down. In case of a fire event, it is dangerous and difficult for the fire brigades to extinguish the fire for they have no possibility to de-energize the PV-plant. The signals of series or parallel arc faults in PV system attenuated by propagation and modified by system noise are non-periodic with low signal to noise ratio. Conventional DC arc frequency domain detections based on Fourier transform are difficult to make effective decisions on DC arc-faults in PV system. This project will focus on some key problems on PV DC arc-faults, such as the signal characteristics of DC arc-faults, physical modeling of arc signal propagating and system noise, combination of time-frequency domain arc signal processing and arc-fault circuit interrupter (AFCI) protection and so on. Experiment tests will be carried on conventional polycrystalline silicon PV modules. Through theory and principle analyzing, modeling, numerical analysis and simulating experimental test, a novel fault detection technique of PV DC arc faults based on time-frequency domain analysis and feature extraction, as well as an method of AFCI protection will be developed. This will provide reliable theoretical and technical support for DC arc-fault detection and AFCI protection in PV system.

随着光伏装机容量不断增加，光伏系统DC电弧故障引发的火灾频繁出现，一旦引起火灾，灭火过程非常困难和危险，但是光伏系统DC电弧类型多样，电弧信号受传播路线衰减及系统噪声影响，具有非周期、低信噪比特性，常规FFT频域检测法难以对其进行有效检测和评判。本项目将从DC电弧故障信号特征、信号传播特性及系统噪声物理建模、结合时频域信号处理方法，电弧故障断路保护（AFCI）等几个突出问题入手，选择多晶硅作为实验对象，通过理论与机理分析、建模、数值分析和模拟实验等研究手段，发展一种基于时频分析及特征提取的检测和电弧故障保护方法，为光伏系统DC电弧检测及电弧故障保护提供可靠理论和技术支撑。

建立了大型光伏电站高频等效模型，综合了光伏组件动态等效模型、线缆高频等效模型及各类寄生参数，研究了故障电弧信号在系统中传播行为，建立了模拟各类故障电弧噪声信号模型。提出了基于故障电弧静态工作点和电阻特性的电弧电流噪声强度模型；提出了故障电弧信号传播行为高频模型；研究了不同拓扑并网逆变器高频噪声信号模型及与故障电弧信号相互影响机理。. 提出了直流电弧综合检测方法，采用了建立时频域特征平面,集合经验模态分解和模糊C均值聚类组合,小波-奇异值分解消噪方法等综合方法，解决了弱电弧、强干扰环境电弧检测问题。提出了一种适用光伏系统故障电弧燃烧的临界条件分析方法；提出了一种基于特征平面电弧检测方法；提出了一种基于集合经验模态分解和模糊C均值聚类的组合故障检测方法；提出了小波-奇异值分解弱电弧检测方法，提出基于SSTDR故障电弧在线检测和定位方法；提出基于快速傅里叶变换（FFT）和BP神经网络相结合的电弧检测方法. 针对光伏阵列中串联型电弧和并联型电弧需要采用不同的电路保护策略，研究了电弧类型辨识方法，提出基于后断前检法采集断点前直流母线电压信号，提取信号的时频域特征来辨识光伏系统中的串行电弧和并行电弧；针对目前国内外缺乏光伏系统故障电弧危害分析方法，提出一种分析光伏系统发生电弧故障的临界熄灭条件的方法，根据熄灭条件确定不同类型电弧的危害程度模型。. 提出了光伏电缆、接地故障亚健康主动检测和定位方法，难点是亚健康故障特征弱，常规手段难以检测，采用了扩展频谱时域反射、高频信号注入等方法，解决了光伏系统形成强电弧危害之前对故障的检测和定位，消除电弧故障隐患。提出了一种基于扩展频谱时域反射法的接地故障检测方法；提出了一种时域反射法直流电缆故障检测和定位方法；提出了一种基于高频信号注入的光伏系统接地故障检测及定位方法；此提出了一种基于SSTDR的母线损伤在线检测方法。