1. 北京化工大学信息科学与技术学院,北京,100029
2. 东南大学复杂工程系统测量与控制教育部重点实验室,江苏,南京,210096
3. 中国科学院理化技术研究所,北京,100190
4. 清华大学自动化系,北京,100084
5. 北京化工大学信息科学与技术学院,北京,100029
6. 东南大学复杂工程系统测量与控制教育部重点实验室,江苏,南京,210096
7. 中国科学院理化技术研究所,北京,100190
8. 清华大学自动化系,北京,100084
纸质出版:2014
移动端阅览
曹政才, 李博, 刘民, 等. 基于动态神经网络的质子交换膜燃料电池建模方法[J]. 电子学报, 2014,42(1):102-106.
CAO Zheng-cai, LI Bo, LIU Min, et al. Approach to Proton Exchange Membrane Fuel Cell Modeling Based on Dynamic Neural Networks[J]. Acta Electronica Sinica, 2014, 42(1): 102-106.
曹政才, 李博, 刘民, 等. 基于动态神经网络的质子交换膜燃料电池建模方法[J]. 电子学报, 2014,42(1):102-106. DOI: 10.3969/j.issn.0372-2112.2014.01.016.
CAO Zheng-cai, LI Bo, LIU Min, et al. Approach to Proton Exchange Membrane Fuel Cell Modeling Based on Dynamic Neural Networks[J]. Acta Electronica Sinica, 2014, 42(1): 102-106. DOI: 10.3969/j.issn.0372-2112.2014.01.016.
针对现有质子交换膜燃料电池(Proton Exchange Membrane Fuel Cell,PEMFC)模型逼近能力不足、自适应性差的问题,提出一种基于动态神经网络的PEMFC建模方法.该方法引入神经网络输出敏感度作为隐含层结构合理性判别依据,根据敏感度分析结果选择采用相应的神经元修改算法调整隐含层结构,使隐含层神经元数目根据燃料电池数据处理需求动态变化,实现模型结构与参数的双重优化.以某型双系统燃料电池测试平台实际运行数据为例进行验证,结果表明构建的PEMFC动态神经网络模型比传统模型的网络规模小、拟合精度高、收敛速度快,适用于工程化仿真应用.
An innovative approach of proton exchange membrane fuel cell (PEMFC) modeling based on dynamic neural networks is proposed to improve approximating and self-adaptive ability of the existing PEMFC models.To evaluate the rationality of networks structure
sensitivity analysis (SA) of the model output was introduced.The hidden nodes were pruned or inserted according to the result of SA to optimize the networks structure and parameters
so that the networks could adapt the PEMFC data processing automatically.The approach was validated using operation data from a commercial dual-system fuel cell test platform.The result shows the proposed PEMFC model with more compact structure
higher accuracy and faster convergence rate compared with the common models
have the capability to be applied to engineering simulation applications.
0
浏览量
2
下载量
1
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621