摘要
给出了一种用于直接甲醇燃料电池(DMFC)温度响应的人工智能建模方法。采用BP网络、Elman神经网络与自适应神经模糊推理系统(ANFIS)避免了精确数学模型的复杂性。详细描述了辨识结构、算法和模型训练方案。通过不同模型之间的仿真对比,得出了三种网络的最优结构参数。仿真结果满足精度要求,得出ANFIS模型性能最优。在ANFIS模型基础上对DMFC温度响应特性作了简要分析。通过仿真模型可快速得到其输入输出特性,使DMFC温度响应的在线预测与控制成为可能。
An artificial intelligence modeling approach was proposed for the temperature response of the direct methanol fuel cell （DMFC）. The complexity of an accurate mathematical model was avoided by using BP network, Elman neural network and adaptive neuro-fuzzy inference system （ANFIS）. The identification structure, algorithms and model training program were described in detail. The optimal structure parameters oft_he three network types were given through the simulation comparisons between different models. The simulation results satisfy the precision demand, and the performance of ANFIS model is the best. The temperature characteristics of DMFC were briefly analyzed based on the ANFIS model. The input-output characteristics could be quickly gotten by the simulation model, which makes it possible to predict and control the temperature response of DMFC online.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第7期1675-1682,共8页
Journal of System Simulation
基金
Science and Technology Innovation Talents Scheme of Henan Province(084200510009)
