采用ELM和优化电压传感器布局的光伏阵列故障检测与区域定位

为快速检测并定位光伏阵列中出现的故障，提出一种新的传感器布局策略，通过优化电压传感器的位置减少电压传感器的数量，同时显化故障的特征.然后，将定位问题转化为分类问题，选用极限学习机(ELM),将最大功率点处的电压数据作为输入训练分类模型.结合实验室屋顶光伏并网发电平台获取的故障数据，对健康状态和所设置的3种故障状态下细化的故障共18种类别，进行分类模型的建立与测试.实验表明，应用本模型故障检测与区域定位的精确率达99.52%,优于所对比的支持向量机、多层感知机网络和随机森林的诊断结果.

Fault detection and location for photovoltaic array using an optimal voltage sensor placement and extreme learning machine

To quickly detect and locate the faults in the photovoltaic array, a new sensor placement strategy is proposed, which reduces the number of voltage sensors by optimizing the location of voltage sensors and also makes the characteristics of the faults more markedly. Then, this research transforms the fault location into a classification issue. According to the characteristics of faults, the extreme learning machine (ELM) is applied to creat the classification model based on the voltage data at the maximum power point. Based on the rooftop grid-connected photovoltaic system in our laboratory, the model is trained to diagnose and locate for 18 types of faults including the fault-free condition and the studied three fault states. The experimental result shows that the classification precision of this model are 99.52% , which is superior to those of support vector machine (SVM), multilayer perceptron (MLP) network, and random forest (RF) .