武汉理工大学学报(信息与管理工程版)
武漢理工大學學報(信息與管理工程版)
무한리공대학학보(신식여관리공정판)
JOURNAL OF WUHAN AUTOMOTIVE POLYTECHNIC UNIVERSITY
2014年
6期
790-793
,共4页
磷酸铁锂电池%SOC估算%BP神经网络
燐痠鐵鋰電池%SOC估算%BP神經網絡
린산철리전지%SOC고산%BP신경망락
lithium iron phosphate battery%SOC estimation%BP neural network
以磷酸铁锂电池为研究对象,根据电池的充放电特性,在Matlab上建立合适的神经网络模型,提出组合训练法,通过大量试验,在比较了10多种训练函数的基础上,得出效果比较好的4种训练函数,兼顾估算精度和训练时间,找出了网络隐含层较优节点数为20,隐含层和输出层的传递函数分别为trainsig和pure-lin。训练结果表明,所建立的BP神经网络模型估算精度高,普适性好。
以燐痠鐵鋰電池為研究對象,根據電池的充放電特性,在Matlab上建立閤適的神經網絡模型,提齣組閤訓練法,通過大量試驗,在比較瞭10多種訓練函數的基礎上,得齣效果比較好的4種訓練函數,兼顧估算精度和訓練時間,找齣瞭網絡隱含層較優節點數為20,隱含層和輸齣層的傳遞函數分彆為trainsig和pure-lin。訓練結果錶明,所建立的BP神經網絡模型估算精度高,普適性好。
이린산철리전지위연구대상,근거전지적충방전특성,재Matlab상건립합괄적신경망락모형,제출조합훈련법,통과대량시험,재비교료10다충훈련함수적기출상,득출효과비교호적4충훈련함수,겸고고산정도화훈련시간,조출료망락은함층교우절점수위20,은함층화수출층적전체함수분별위trainsig화pure-lin。훈련결과표명,소건립적BP신경망락모형고산정도고,보괄성호。
The lithium iron phosphate battery was studied.The charge-discharge characteristics of the battery were investi-gated.Then the appropriate neural network was built in MATLAB.A combined training method was proposed.After comparison of more than ten train functions, four functions'performance were turn out to be outstanding.When both estimation accuracy and training time were considered, the appropriate number of hidden nodes was 20.Two transfer functions are tansig and purelin.The network trained by sample data was verified to be universal, and high estimation was achieved.