基于深度置信网络的光伏发电阵列的故障诊断方法
Fault diagnosis methods of photovoltaic power array based on deep belief network
彭辉 1田程程 1郑宇锋 2黄婧柠1
作者信息
- 1. 武汉大学 电气与自动化学院,武汉 430072
- 2. 海军工程大学 电磁能技术全国重点实验室,武汉 430033
- 折叠
摘要
为了对光伏发电阵列的故障进行及时准确的诊断,首先以在 Matlab/Simulink中搭建的光伏阵列输出特性仿真模型为基础,采用深度置信网络作为光伏阵列故障诊断算法,并通过蝙蝠算法对网络各隐含层神经元的数量进行优化;然后,以蝙蝠算法优化后的深度置信网络(bat algorithm-deep belief network,BA-DBN)作为故障诊断模型,分别采集不同运行工况下的光伏阵列输出特性四参数,并将其归一化后作为特征样本输入BA-DBN故障诊断模型,实现了对光伏阵列的故障诊断.仿真结果表明:所提出的BA-DBN 算法在光伏阵列故障诊断应用中的准确率显著高于KNN、BPNN和原始DBN算法,更加适用于光伏阵列故障诊断,具有更优的分类效果.
Abstract
In order to timely and accurately diagnose the fault of photovoltaic power array,firstly,the deep confidence network was used as the fault diagnosis algorithm of photovoltaic array,which was based on the simulation model of photovoltaic array output characteristics built in Matlab/Simulink.Then,the number of neurons in each hidden layer of the network was optimized by bat algorithm,and the optimized bat algorithm-deep belief network(BA-DBN)was used as the fault diagnosis model.Four parameters of photovoltaic array output characteristics under different operating conditions were obtained as characteristics samples,and then were input into the BA-DBN fault diagnosis model after normalization to realize the fault diagnosis of photovoltaic array.It is proved that the correct ratio of grid-connected photovoltaic power generation system based on the proposed BA-DBN algorithm is much higher than KNN,BPNN,and the original DBN algorithms.It is more suitable for photovoltaic array fault diagnosis and has better classifieation performance.
关键词
光伏阵列/故障诊断/深度置信网络/蝙蝠算法Key words
PV array/fault diagnosis/deep belief network/bat algorithm引用本文复制引用
基金项目
国家自然科学基金(92266201)
中国博士后科学基金(2019T120972)
出版年
2024
NETL
NSTL
万方数据