不确定条件下车用燃料电池故障预测

As the core component of fuel cell vechiles, the healthy operation of fuel cell is the key to ensure fuel cell vehicles safe driving. Membrane drying and water flooding are the most common faults, which have a great influence on the health of fuel cell. The current research on the membrane drying and water flooding is mainly concentrated in the fault diagnostics.However, few of them addresses the prediction of the remaining life, especially considering the fuel cell aging and the operating parameters (temperature, humidity, etc.) uncertainty. Based on the research in the fuel cell system, the applicant proposes an improved method which combines an age-dependent Hidden semi-Markov Model(HSMM)and a dynamic failure rate model. The study mainly includes as follows. Firstly introduce an aging factor to the HSMM to describe the uncertainty of fuel cell aging. Then the maximum likelihood function is used to recognize the symptoms of the fuel cell membrane drying and water flooding.Lastly considering the uncertainty of the operating conditions in the fuel cell, built a dynamic failure rate model to predict the remaining life of the fuel cell. This project provides an analysis method and scientific basis for the optimal maintenance of fuel cell vehicles.

作为燃料电池汽车的核心部件，车用燃料电池健康运行是确保燃料电池汽车安全行驶的关键。膜干和水淹是车用燃料电池最常见的两种故障，也是影响其健康的重要因素。当前对膜干和水淹故障研究主要集中在诊断故障是否发生，缺乏能有效预测何时发生故障的方法。本项目在申请者近年来从事燃料电池控制与优化研究基础上，利用基于老化因子的隐式半马尔科夫模型与动态失效率模型相结合的方法，对上述两种不确定条件下的车用燃料电池膜干和水淹故障进行预测研究。研究内容主要包括：考虑燃料电池老化不确定性影响，建立基于老化因子的隐式半马尔科夫故障诊断及预测模型；基此利用极大似然函数估计方法对膜干和水淹故障早期衰退进行诊断；进一步针对燃料电池操作条件的不确定性影响，构建动态失效率模型，预测发生膜干或水淹故障的有效剩余寿命。本项目研究为燃料电池汽车最优维护提供分析方法和科学依据。

剩余寿命预测算法的研究是提高燃料电池可靠性和持久性的一个重要问题，然而目前对其研究仍处在起步阶段。本课题提出隐式半马尔科夫模型（HSMM）与经验模型相结合的混合模型，预测燃料电池剩余寿命。基于Bp神经网络探讨了质子交换膜燃料电池（PEMFC）水淹和膜干的故障诊断，利用LS-SVM分类器，辨识了固体氧化物燃料电池（SOFC）阳极中毒及阴极中毒故障。采用HSMM方法预测了SOFC阳极中毒及阴极中毒下的剩余寿命。针对燃料电池的不确定性，将HSMM与经验模型相结合，预测了不确定情况下燃料电池的剩余寿命。对提出的预测方法进行了仿真与分析，结果表明能有效的预测燃料电池剩余寿命，该方法能有效保护燃料电池，提高系统寿命。