准确估计锂离子电池的健康状态(state of health,SOH)对保证电池的安全使用具有十分重要的意义。为了提高SOH估计精度,提出了一种融合卷积神经网络(convolutional neural network,CNN)和Transformer的锂离子电池SOH估计方法。首先,分析...准确估计锂离子电池的健康状态(state of health,SOH)对保证电池的安全使用具有十分重要的意义。为了提高SOH估计精度,提出了一种融合卷积神经网络(convolutional neural network,CNN)和Transformer的锂离子电池SOH估计方法。首先,分析了牛津实验室测试得到的锂离子电池充放电循环数据,提取了部分等间隔电压对应的充电时间序列作为表征电池老化的健康特征,并利用Pearson相关系数法定量分析了健康特征与SOH直接的相关性。然后,将具有局部特征提取能力的CNN与具有自注意力全局特征提取能力的Transformer相结合进行SOH估计。为了进一步提高估算精度,采用贝叶斯优化算法对CNN-Transformer全局超参数进行寻优,得到最优模型参数组合,提高了模型计算速度和SOH估算精度。在8个电池中进行交叉验证,结果表明:所提出的方法可以保证SOH估算最大误差、均方根误差和平均绝对误差分别小于1.5%、0.75%、0.63%。并将提出的方法与4种传统深度学习算法进行比较分析,发现该方法具有更好的估算精度和泛化能力。展开更多
锂离子电池容量的精确预测有利于提升电池的使用安全和避免电池滥用,受其复杂的内部电化学反应和外部使用条件影响,对其老化后的容量精确预测一直是电池管理系统的难点之一。为实现锂电池全服役周期内容量的高效准确预测,本文提出了一...锂离子电池容量的精确预测有利于提升电池的使用安全和避免电池滥用,受其复杂的内部电化学反应和外部使用条件影响,对其老化后的容量精确预测一直是电池管理系统的难点之一。为实现锂电池全服役周期内容量的高效准确预测,本文提出了一种基于遗传算法优化的Elman神经网络(GA-Elman)电池容量预测模型。首先选择电池不同循环下的放电容量增量、内阻以及温度数据作为有效表征电池老化和容量衰减规律的特征量,其次运用主成分分析算法对特征量进行降维以降低训练量数据维度,然后基于Elman神经网络构建电池容量预测模型,并引入遗传算法优化Elman神经网络的权值和阈值,实现对电池容量的高效精确预测,最后在不同电池上对该模型进行了验证。验证结果表明:与传统Elman神经网络和长短期记忆神经网络(long and short term memory neural network,LSTM NN)预测模型相比,GA-Elman神经网络预测模型有更好的预测精度和更高的运算效率。在不同电池上该模型预测结果的最大平均绝对误差为0.92%,最大均方根误差为1.02%,最小拟合系数为0.9679,表明该模型可以精确预测锂电池衰退过程中的容量并且对不同电池有较强的适应性。展开更多
Rosa rugosa Thunb.is recognized as both medicine and edible in China.The article investigated the antitumor activity of rose flavonoids.Water-extracted rose flavonoids(RFW)and ethanol-extracted rose flavonoids(RFE)wer...Rosa rugosa Thunb.is recognized as both medicine and edible in China.The article investigated the antitumor activity of rose flavonoids.Water-extracted rose flavonoids(RFW)and ethanol-extracted rose flavonoids(RFE)were achieved by extracting with distilled water and 70%ethanol,respectively.The effects of the two extracts on proliferation inhibition,apoptosis inducement and metastasis prevention of human HepG2 hepatocellular carcinoma cell lines were tested,via optical/fluorescence microscopy,MTT detection,Transwell assay,flow cytometry and Western blot,etc.The results indicated that rose flavonoids at low concentration(10-40μg/mL)had a better inhibitory effect on migration and invasion of HepG2 cells in a dose-dependent manner,while rose flavonoids at high concentration(80-160μg/mL)could induce apoptosis of HepG2 cells by up-regulating the expression of pro-apoptotic proteins p53 and Bax,and down-regulating the expression of anti-apoptotic proteins Bcl-2,leading to the functioning of caspase-3 and caspase-9.The effect of RFE at the same concentration was significantly better than that of RFW.Conclusion,this study found that rose flavonoids had a certain inhibitory effect on proliferation and metastasis of human liver cancer cells HepG2,indicating the application of rose flavonoids in preventing and treating of liver cancer.展开更多
文摘准确估计锂离子电池的健康状态(state of health,SOH)对保证电池的安全使用具有十分重要的意义。为了提高SOH估计精度,提出了一种融合卷积神经网络(convolutional neural network,CNN)和Transformer的锂离子电池SOH估计方法。首先,分析了牛津实验室测试得到的锂离子电池充放电循环数据,提取了部分等间隔电压对应的充电时间序列作为表征电池老化的健康特征,并利用Pearson相关系数法定量分析了健康特征与SOH直接的相关性。然后,将具有局部特征提取能力的CNN与具有自注意力全局特征提取能力的Transformer相结合进行SOH估计。为了进一步提高估算精度,采用贝叶斯优化算法对CNN-Transformer全局超参数进行寻优,得到最优模型参数组合,提高了模型计算速度和SOH估算精度。在8个电池中进行交叉验证,结果表明:所提出的方法可以保证SOH估算最大误差、均方根误差和平均绝对误差分别小于1.5%、0.75%、0.63%。并将提出的方法与4种传统深度学习算法进行比较分析,发现该方法具有更好的估算精度和泛化能力。
文摘锂离子电池容量的精确预测有利于提升电池的使用安全和避免电池滥用,受其复杂的内部电化学反应和外部使用条件影响,对其老化后的容量精确预测一直是电池管理系统的难点之一。为实现锂电池全服役周期内容量的高效准确预测,本文提出了一种基于遗传算法优化的Elman神经网络(GA-Elman)电池容量预测模型。首先选择电池不同循环下的放电容量增量、内阻以及温度数据作为有效表征电池老化和容量衰减规律的特征量,其次运用主成分分析算法对特征量进行降维以降低训练量数据维度,然后基于Elman神经网络构建电池容量预测模型,并引入遗传算法优化Elman神经网络的权值和阈值,实现对电池容量的高效精确预测,最后在不同电池上对该模型进行了验证。验证结果表明:与传统Elman神经网络和长短期记忆神经网络(long and short term memory neural network,LSTM NN)预测模型相比,GA-Elman神经网络预测模型有更好的预测精度和更高的运算效率。在不同电池上该模型预测结果的最大平均绝对误差为0.92%,最大均方根误差为1.02%,最小拟合系数为0.9679,表明该模型可以精确预测锂电池衰退过程中的容量并且对不同电池有较强的适应性。
基金Supported by Grants from the National High-tech R and D Pro-gram No.2012AA020206the Key Project for the Infectious Diseases No.2012ZX10002-017 and No.2013ZX10002009-001-004+2 种基金the State Key Projects for Basic Research No.2011CB910703the National Natural Science Foundation No.81372591,and No.81321091 of Chinathe Center for Marine Medicine and Rescue of Tsinghua University
文摘AIM: To investigate the expression of key biomarkers in hepatoma cell lines, tumor cells from patients’ blood samples, and tumor tissues.
基金supported by the natural science foundation of Shandong province ZR2017BH053the youth doctor cooperation foundation of Qilu University of Technology(Shandong Academy of Sciences)2017BSH2017。
文摘Rosa rugosa Thunb.is recognized as both medicine and edible in China.The article investigated the antitumor activity of rose flavonoids.Water-extracted rose flavonoids(RFW)and ethanol-extracted rose flavonoids(RFE)were achieved by extracting with distilled water and 70%ethanol,respectively.The effects of the two extracts on proliferation inhibition,apoptosis inducement and metastasis prevention of human HepG2 hepatocellular carcinoma cell lines were tested,via optical/fluorescence microscopy,MTT detection,Transwell assay,flow cytometry and Western blot,etc.The results indicated that rose flavonoids at low concentration(10-40μg/mL)had a better inhibitory effect on migration and invasion of HepG2 cells in a dose-dependent manner,while rose flavonoids at high concentration(80-160μg/mL)could induce apoptosis of HepG2 cells by up-regulating the expression of pro-apoptotic proteins p53 and Bax,and down-regulating the expression of anti-apoptotic proteins Bcl-2,leading to the functioning of caspase-3 and caspase-9.The effect of RFE at the same concentration was significantly better than that of RFW.Conclusion,this study found that rose flavonoids had a certain inhibitory effect on proliferation and metastasis of human liver cancer cells HepG2,indicating the application of rose flavonoids in preventing and treating of liver cancer.
