燃料电池预测模型输出结果统计分析
Statistical analysis of the output results of fuel cell prediction model
鲁源博 1侯永平 1焦道宽 2王要娟3
作者信息
- 1. 同济大学汽车学院,上海 201804
- 2. 中汽研新能源汽车检验中心(天津)有限公司,天津 300300
- 3. 上海机动车检测认证技术研究中心有限公司,上海 201805
- 折叠
摘要
基于神经网络算法建立的燃料电池寿命预测模型,输出结果都因随机性存在不确定的问题,即每次预测的输出结果都不同.针对此问题,基于长短时记忆(LSTM)神经网络算法建立燃料电池寿命预测模型,多次运行试验样本数据,利用统计学方法对输出结果的分布规律进行统计特性分析,发现基于LSTM神经网络的寿命预测模型,输出结果符合正态分布规律.根据此结论,可采用多次平均结果作为燃料电池寿命预测模型的输出结果,以提升输出结果的预测精度及稳定性.
Abstract
Due to the randomness,fuel cell lifetime prediction models based on neural network algorithms have uncertainty in their output results,meaning that the output is different with each prediction.To address this issue,a fuel cell lifetime prediction model based on the long short-term memory(LSTM)neural network algorithm is established.This model is run multiple times on the experimental sample data,statistical methods are used to analyze the statistical characteristics of the distribution of output results.It is found that the output results of the LSTM neural network-based lifetime prediction model follow a normal distribution pattern.Based on this conclusion,the average of multiple results can be used as the output of the fuel cell lifetime prediction model,thereby improving the prediction accuracy and stability of the output results.
关键词
燃料电池/寿命预测模型/正态分布检验/长短时记忆(LSTM)神经网络/统计特性Key words
fuel cell/lifetime prediction model/normal distribution test/long short-term memory(LSTM)neural network/statistical characterization引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据