基于MVMD和ISCSO-HKELM的质子交换膜燃料电池故障诊断  

Fault diagnosis of proton exchange membrane fuel cells based on MVMD and ISCSO-HKELM

作　　者：杜董生 连贺 邓祥帅 任一鸣 赵哲敏 DU Dongsheng;LIAN He;DENG Xiangshuai;REN Yiming;ZHAO Zhemin(School of Automation,Huaiyin Institute of Technology,Huai'an 223003,China)

机构地区：[1]淮阴工学院自动化学院,江苏淮安223003

出　　处：《综合智慧能源》2024年第12期17-28,共12页Integrated Intelligent Energy

基　　金：国家自然科学基金项目(62173159)。

摘　　要：针对质子交换膜燃料电池(PEMFC)系统的原始信号受到高温和强背景噪声影响导致故障诊断准确率较低的问题,提出一种基于多元变分模态分解(MVMD)和改进沙丘猫优化算法(ISCSO)优化混合核极限学习机(HKELM)的PEMFC故障诊断模型。通过小波硬阈值(WHTD)对PEMFC的原始信号进行去噪,利用MVMD将去噪后信号进行模态分解进而得到一系列本征模态函数(IMF),利用方差贡献率、相关系数和信息熵筛选出最优的IMF进行信号重构。通过逻辑(Logistic)映射、透镜成像折射反向学习(ROBL)、非线性动态因子和黄金正弦策略改进沙丘猫算法(SCSO),得到ISCSO。利用ISCSO对HKELM进行优化,并基于改进后的ISCSO-HKELM对重构信号进行特征提取进而实现故障诊断。将所提出的WHTD-MVMD-ISCSO-HKELM故障诊断模型与其他算法进行对比验证,试验结果表明,所提方法能够明显提升故障诊断的准确率,具有一定的可行性和优越性。To address the issue of low fault diagnostic accuracy in proton exchange membrane fuel cell(PEMFC)systems caused by high temperatures and strong background noise in raw signals,a fault diagnosis model based on multivariate variational mode decomposition(MVMD)and improved sand cat swarm optimization(ISCSO)algorithm for optimizing the hybrid kernel extreme learning machine(HKELM)was proposed.Wavelet hard thresholding denoising(WHTD)was applied to filter the raw PEMFC signals,and MVMD was employed to decompose the denoised signals into a series of intrinsic mode functions(IMFs).The optimal IMFs were selected for signal reconstruction based on variance contribution rate,correlation coefficient,and information entropy.The sand cat swarm optimization(SCSO)algorithm was enhanced using logistic mapping,refracted opposition-based learning(ROBL),nonlinear dynamic factors,and golden sine strategy,resulting in an improved ISCSO algorithm.The ISCSO algorithm was then applied to optimize HKELM,and the improved ISCSO-HKELM algorithm was utilized to extract features from the reconstructed signals for fault diagnosis.The proposed WHTD-MVMD-ISCSO-HKELM fault diagnosis model was compared with other algorithms,and the experimental results demonstrated that the proposed approach significantly improved fault diagnostic accuracy,indicating feasibility and superiority.

关 键 词：混合核极限学习机 改进沙丘猫优化算法 多元变分模态分解 质子交换膜燃料电池 故障诊断 

分 类 号：TK01[动力工程及工程热物理] TP181[自动化与计算机技术—控制理论与控制工程]