FEMTO-ST, UMR CNRS 6174, Université Bourgogne Franche-Comté, Belfort/UTBM 90000, France; FCLAB, FR CNRS 3539, Université Bourgogne Franche-Comté, Belfort/UTBM 90000, France.
ISA Trans. 2021 Jul;113:175-184. doi: 10.1016/j.isatra.2020.03.012. Epub 2020 Mar 10.
This paper proposes a novel degradation prognosis of Proton Exchange Membrane Fuel Cell (PEMFC) based on Wavelet Neural Network (WNN) and Cuckoo Search Algorithm (CSA). The proposed method considering the main operating conditions of PEMFC can be applied to the health state prognostic of PEMFC under different conditions. First, the operating data of PEMFC are reconstructed by the locally weighted scatterplot smoothing method to filter noise. Then, the WNN that can analyze the degradation characteristics of PEMFC (global degradation trend and reversible phenomena) is adopted to build the degradation model of PEMFC. Finally, the structure and parameters of WNN are optimized by CSA to improve the accuracy for the degradation prognosis of PEMFC. The optimized degradation prognosis method is used to predict the remaining useful life of PEMFC. The proposed prognostic method is validated by 3 degradation tests of PEMFC under different conditions. The results show that CSA can greatly improve the degradation prognosis accuracy of PEMFC based on WNN. The proposed CSA-WNN can achieve higher precision than other traditional prognostic methods.
本文提出了一种基于小波神经网络（WNN）和布谷鸟搜索算法（CSA）的质子交换膜燃料电池（PEMFC）的新型降解预测方法。所提出的方法考虑了 PEMFC 的主要工作条件，可以应用于不同条件下 PEMFC 的健康状态预测。首先，通过局部加权散点平滑方法对 PEMFC 的工作数据进行重构，以滤除噪声。然后，采用能够分析 PEMFC 降解特性（全局降解趋势和可逆现象）的 WNN 来建立 PEMFC 的降解模型。最后，通过 CSA 优化 WNN 的结构和参数，提高 PEMFC 降解预测的准确性。优化的降解预测方法用于预测 PEMFC 的剩余使用寿命。通过在不同条件下进行的 3 个 PEMFC 降解测试对所提出的预测方法进行验证。结果表明，CSA 可以大大提高基于 WNN 的 PEMFC 降解预测精度。所提出的 CSA-WNN 可以比其他传统预测方法实现更高的精度。
