Remaining Useful Life Prediction for Fuel Cell Based on Support Vector Regression and Grey Wolf Optimizer Algorithm
Autor: | Kui Chen, Salah Laghrouche, Abdesslem Djerdir |
---|---|
Přispěvatelé: | Femto-st, Energie |
Rok vydání: | 2022 |
Předmět: |
[SPI.AUTO] Engineering Sciences [physics]/Automatic
[PHYS.MECA.THER] Physics [physics]/Mechanics [physics]/Thermics [physics.class-ph] Energy Engineering and Power Technology [PHYS.MECA.MEFL] Physics [physics]/Mechanics [physics]/Fluid mechanics [physics.class-ph] Electrical and Electronic Engineering [SPI.NRJ] Engineering Sciences [physics]/Electric power |
Zdroj: | IEEE Transactions on Energy Conversion. 37:778-787 |
ISSN: | 1558-0059 0885-8969 |
DOI: | 10.1109/tec.2021.3121650 |
Popis: | Remaining useful life prediction is an important way to improve thedurability and reduce the cost of the proton exchange membrane fuelcell. This paper presents a novel method to predict the remaininguseful life of proton exchange membrane fuel cell under differentload currents based on support vector regression and grey wolfoptimizer algorithm. The proposed method considers the influence of17 operating conditions and historical voltage. Firstly, themeasured data are reconstructed through robust locally weightedsmoothing method to reduce the calculation amount and filterdisturbances. Then, support vector regression with fewerhyperparameters is used to establish the degradation model.Finally, the hyperparameters of support vector regression areoptimized through grey wolf optimizer algorithm to improve theaccuracy of degradation prediction. The proposed method isvalidated by two degradation experiments under different loadcurrents. The test results show that grey wolf optimizer algorithmcan effectively improve the accuracy of degradation predictionbased on support vector regression. Compared with other methods,the proposed method has the highest accuracy. The proposed methodcan predict the fuel cell degradation with a mean absolutepercentage error of less than 0.3%. The proposed method can predictthe remaining useful life of 492 hours. |
Databáze: | OpenAIRE |
Externí odkaz: |