引用本文: | 刘昕宇,韩莹,陈维荣,李奇,杨哲昊.基于改进SSA-DBN的质子交换膜燃料电池水故障智能分类方法[J].电力自动化设备,2024,44(4):18-24 |
| LIU Xinyu,HAN Ying,CHEN Weirong,LI Qi,YANG Zhehao.Intelligent classification method of water faults for proton exchange membrane fuel cell based on improved SSA-DBN[J].Electric Power Automation Equipment,2024,44(4):18-24 |
|
摘要: |
为了实现质子交换膜燃料电池(PEMFC)系统水故障的高效快速分类,提出了基于改进麻雀搜索算法(SSA)优化深度置信网络(DBN)的PEMFC故障分类方法。采用归一化处理消除故障数据参数之间量纲不同的影响,使用核主成分分析对数据进行故障特征提取,有效地缩减了原始数据维度,降低了运算复杂度,并避免低贡献度数据对故障分类造成干扰。引入柯西-高斯变异策略改进SSA,并利用SSA对DBN进行参数寻优,确定网络结构,通过优化后的DBN实现对PEMFC水故障的快速分类。对3 000组PEMFC水故障数据进行测试,结果表明:所提方法可以快速准确地识别PEMFC的正常状态、膜干故障、水淹故障3种健康状态;总体的分类准确率为98.67%,运算时间为0.89 s,相比支持向量机、概率神经网络方法,所提方法的故障分类精度分别提升了4 %、3.34 %,运算时间分别减少了15.35、0.35 s。 |
关键词: 质子交换膜燃料电池 故障分类 深度置信网络 麻雀搜索算法 核主成分分析 柯西-高斯变异策略 |
DOI:10.16081/j.epae.202310008 |
分类号:TM911.4 |
基金项目:国家自然科学基金资助项目(52007157,52077180);四川省自然科学基金资助项目(2022NSFSC0269) |
|
Intelligent classification method of water faults for proton exchange membrane fuel cell based on improved SSA-DBN |
LIU Xinyu, HAN Ying, CHEN Weirong, LI Qi, YANG Zhehao
|
School of Electrical Engineering, Southwest Jiaotong University, Chengdu 611756, China
|
Abstract: |
In order to realize efficient and rapid classification of water faults in proton exchange membrane fuel cell(PEMFC) system, a fault classification method of PEMFC based on improved sparrow search algorithm(SSA) optimizing deep confidence network(DBN) is proposed. The normalization process is adopted to eliminate the impact of different dimensions among fault data parameters, and the kernel principal component analysis is used to extract fault features from the data, which effectively reduces the dimensions of the original data and the calculation complexity and avoids the interference of low-contribution data on fault classification. The Cauchy-Gauss variation strategy is introduced to improve SSA, the parameters of DBN are optimized by SSA to determine the network structure, and the optimized DBN is used to quickly classify the water faults of PEMFC. Three thousand sets of PEMFC water fault data are tested. The results show that the proposed method can quickly and accurately identify three health states of PEMFC: normal state, membrane dry fault and flooding fault. The overall classification accuracy is 98.67% and the calculation time is 0.89 s. Compared with the support vector machine and the probabilistic neural network method, the fault classification accuracy of the proposed method is increased by 4 % and 3.34 % respectively, and the calculation time is reduced by 15.35 s and 0.35 s respectively. |
Key words: proton exchange membrane fuel cell fault classification deep confidence network sparrow search algorithm kernel principal component analysis Cauchy-Gaussian variation strategy |