针对锂离子电池荷电状态(state of charge,SOC)预测问题,利用长短期记忆(long short-term memory,LSTM)循环神经网络建立电池SOC预测模型。在恒阻放电情况下,将电池输出电流、输出电压和电池表面温度作为模型的主要输入,使用训练样本对...针对锂离子电池荷电状态(state of charge,SOC)预测问题,利用长短期记忆(long short-term memory,LSTM)循环神经网络建立电池SOC预测模型。在恒阻放电情况下,将电池输出电流、输出电压和电池表面温度作为模型的主要输入,使用训练样本对神经网络进行训练,使用验证样本进行验证。结果表明,用该方法进行电池SOC预测时可使最大绝对误差仅为1.96%,均方根误差为0.986%,可行性被验证。分析神经网络隐含层中不同的神经元个数对预测结果的影响,对比不同批大小情况下训练出的神经网络的预测误差。将隐含层分别设置为1至3个LSTM细胞核,得到不同条件下神经网络的预测误差。结果为电池SOC预测的神经网络模型的隐含层神经元个数、批大小和LSTM细胞核个数的设定提供参考。展开更多
为了解决传统安时-开路电压法荷电状态(state of charge,SOC)初值SOC0误差大,忽略了估算过程中温度等影响因素对估算精度的影响等问题,提出了改进的安时积分-开路电压法:根据不同温度、循环使用次数下的实验数据,拟合出SOC与开路电压(op...为了解决传统安时-开路电压法荷电状态(state of charge,SOC)初值SOC0误差大,忽略了估算过程中温度等影响因素对估算精度的影响等问题,提出了改进的安时积分-开路电压法:根据不同温度、循环使用次数下的实验数据,拟合出SOC与开路电压(open circuit voltage,OCV)、温度、使用次数的函数关系,从而获取准确的SOC0;结合实验分析温度、放电倍率、使用次数对于安时积分的影响,并对其进行修正和优化。实验表明,改进的安时-开路电压法可将估算精度提高至97%。展开更多
Defects in the function and development of GABAergic interneurons have been linked to psychiatric disorders, so preservation of these interneurons in brain slices is important for successful electrophysiological recor...Defects in the function and development of GABAergic interneurons have been linked to psychiatric disorders, so preservation of these interneurons in brain slices is important for successful electrophysiological recording in various ex vivo methods. However, it is difficult to maintain the activity and morphology of neurons in slices from mice of 〉30 days old. Here we evaluated the N-methyI-D-glucamine (NMDG)- based artificial cerebrospinal fluid (aCSF) method for the preservation of interneurons in slices from mice of up to -6 months old and discussed the steps that may affect their quality during slicing. We found that the NMDG-aCSF method rescued more cells than sucrose-aCSF and successfully preserved different types of interneurons including parvalbumin- and somatostatin-positive interneurons. In addition, both the chemical and electrical synaptic signaling of interneurons were maintained. These results demonstrate that the NMDG-aCSF method is suitable for the preservation of interneurons, especially in studies of gap junctions.展开更多
文摘针对锂离子电池荷电状态(state of charge,SOC)预测问题,利用长短期记忆(long short-term memory,LSTM)循环神经网络建立电池SOC预测模型。在恒阻放电情况下,将电池输出电流、输出电压和电池表面温度作为模型的主要输入,使用训练样本对神经网络进行训练,使用验证样本进行验证。结果表明,用该方法进行电池SOC预测时可使最大绝对误差仅为1.96%,均方根误差为0.986%,可行性被验证。分析神经网络隐含层中不同的神经元个数对预测结果的影响,对比不同批大小情况下训练出的神经网络的预测误差。将隐含层分别设置为1至3个LSTM细胞核,得到不同条件下神经网络的预测误差。结果为电池SOC预测的神经网络模型的隐含层神经元个数、批大小和LSTM细胞核个数的设定提供参考。
文摘为了解决传统安时-开路电压法荷电状态(state of charge,SOC)初值SOC0误差大,忽略了估算过程中温度等影响因素对估算精度的影响等问题,提出了改进的安时积分-开路电压法:根据不同温度、循环使用次数下的实验数据,拟合出SOC与开路电压(open circuit voltage,OCV)、温度、使用次数的函数关系,从而获取准确的SOC0;结合实验分析温度、放电倍率、使用次数对于安时积分的影响,并对其进行修正和优化。实验表明,改进的安时-开路电压法可将估算精度提高至97%。
基金supported by grants from National Funds for Distinguished Young Scientists of China (81225007)the Funds for Creative Research Groups of China (81221003)
文摘Defects in the function and development of GABAergic interneurons have been linked to psychiatric disorders, so preservation of these interneurons in brain slices is important for successful electrophysiological recording in various ex vivo methods. However, it is difficult to maintain the activity and morphology of neurons in slices from mice of 〉30 days old. Here we evaluated the N-methyI-D-glucamine (NMDG)- based artificial cerebrospinal fluid (aCSF) method for the preservation of interneurons in slices from mice of up to -6 months old and discussed the steps that may affect their quality during slicing. We found that the NMDG-aCSF method rescued more cells than sucrose-aCSF and successfully preserved different types of interneurons including parvalbumin- and somatostatin-positive interneurons. In addition, both the chemical and electrical synaptic signaling of interneurons were maintained. These results demonstrate that the NMDG-aCSF method is suitable for the preservation of interneurons, especially in studies of gap junctions.