【摘要】 Kriging模型具有良好的函数逼近能力，能同时给出预测值和误差估计。引入Kriging模型建立固体氧化物燃料电池(SOFC)代理模型，可在准确预测电池性能的同时降低在线计算成本。为构建SOFC变工况操作下的数据集，建立了三维多物理场仿真模型，数值模拟结果与实验数据吻合。利用随机采样和数值模拟生成样本空间，比较了七种采用不同相关函数的Kriging模型，分析了样本数量对模型准确性的影响以及小样本情况下模型的鲁棒性。结果表明，采用高斯相关函数的Kriging模型在SOFC性能预测方面具有较为优越的综合表现，并给出了平衡计算精度和时间成本的样本数量参考值，具有一定实用价值。更多还原

【Abstract】 Kriging model has a good ability to approximate nonlinear functions, where the predicted value and error estimation can be supplied simultaneously. Therefore, Kriging models were introduced to construct the surrogate model of solid oxide fuel cells(SOFC), which can accurately predict cell performance and reduce online computing costs. A three-dimensional multi-physics simulation model was developed to obtain the sample set. The numerical results show good agreement with experimental data. By using random sampling and numerical simulation to generate sample space, seven Kriging surrogates with different correlation models were compared. The influence of sample size on the model accuracy and the model robustness were analyzed. The results show that the Kriging surrogate with the Gaussian correlation model has superior and comprehensive performance. The reference value of sample size which can balance accuracy and time cost is also given. These results have certain practical value in the surrogate modeling of SOFC.更多还原