摘要
光伏阵列由大量串联和并联的光伏组件组成,使得在某一(些)光伏组件出现故障时,对故障点的精准定位较难,针对此问题,本文提出了基于高斯过程回归机制的光伏阵列故障定位方法。首先,采用二进制码表示光伏阵列工作状态,将各二进制码转换为整数,构建了基于整数回归的故障定位模型;然后,采用高斯过程对转换后的光伏阵列工作状态进行拟合;最后,根据当前光伏阵列输出电压、电流等信号,实现对故障诊断和故障点定位。将所提算法应用于2×4光伏阵列系统,分别针对一个、两个、三个、四个光伏组件出现故障,以及前三种组合状态的情况进行实验,并与常用BP神经网络故障定位方法进行比较。从实验结果可以看出本文所提故障定位方法优于BP神经网络;高斯过程对一个、两个及三个光伏组件出现故障的情况能够进行很好的定位,但是对四个组件同时出现故障的情况存在一定的定位误差。
Photovoltaic （PV） array is often composed with a lot of photovoltaic modules in series or in parallel, which make it difficult to obtain the accurate fault locations when some PV modules are failure. To well dispose this problem, a novel strategy of fault location by use of Gaussian process（GP） regression is presented in this paper. First, binary strings are designed to represent the states of the PV array, and then a regression model of the fault location by converting the binary strings into corresponding integers is presented. Based on the voltage and current as well as the converted integers, the Gaussian process is trained. With newly measured voltage and current, the current state of the PV array is diagnosed by using the trained Gaussian Process. The proposed algorithm is applied to a 2 × 4 PV array and verified via four different fault states, i.e., one module, two modules, three modules and four modules are failure. By comparing with the common used BP network in fault diagnosis and location, the effectiveness of our algorithm is empirically demonstrated. The results show that the Gaussian Process based method outperforms BP network one in more accurately locating the faults. Furthermore, the proposed algorithm is more benefit for the first three fault states than the one that four modules are failure.
出处
《电工技术学报》
EI
CSCD
北大核心
2013年第6期249-256,共8页
Transactions of China Electrotechnical Society
基金
徐州市科技计划资助项目(XJ11B010)
