锂电池健康状态(state of health, SOH)可表征锂电池的老化状态。为准确评估SOH,首先,提取充电阶段的电流、电压、IC曲线中的6个特征,为了提高输入特征的质量,采用核主成分分析(kernel principal component analysis, KPCA)结合Spearma...锂电池健康状态(state of health, SOH)可表征锂电池的老化状态。为准确评估SOH,首先,提取充电阶段的电流、电压、IC曲线中的6个特征,为了提高输入特征的质量,采用核主成分分析(kernel principal component analysis, KPCA)结合Spearman相关性分析,消除多维特征的冗余性获取输入特征的关键信息。其次,为了降低模型复杂度,将全连接层代替Transformer解码器,并利用牛顿-拉夫逊优化算法(Newton-Raphson-based optimization algorithm, NRBO)对模型的超参数寻优,提高预测精度。最后,利用公开数据集不同训练比例划分验证方法的有效性,并采用不同电池交叉验证与灰狼优化算法(gray wolf optimization algorithm, GWO)和鲸鱼优化算法(whale optimization algorithm, WOA)进行比较,结果表明:所提方法在精度和计算耗时方面均优于其他2种算法。展开更多
Steady-state heat conduction problems arisen in connection with various physical and engineering problems where the functions satisfy a given partial differential equation and particular boundary conditions, have attr...Steady-state heat conduction problems arisen in connection with various physical and engineering problems where the functions satisfy a given partial differential equation and particular boundary conditions, have attracted much attention and research recently. These problems are independent of time and involve only space coordinates, as in Poisson＇s equation or the Laplace equation with Dirichlet, Neuman, or mixed conditions. When the problems are too complex, it is difficult to find an analytical solution, the only choice left is an approximate numerical solution. This paper deals with the numerical solution of three-dimensional steady-state heat conduction problems using the meshless reproducing kernel particle method （RKPM）. A variational method is used to obtain the discrete equations. The essential boundary conditions are enforced by the penalty method. The effectiveness of RKPM for three-dimensional steady-state heat conduction problems is investigated by two numerical examples.展开更多
锂电池健康状态(state of health, SOH)的退化过程在一定程度上是一个非平稳随机过程,使得当前多数点估计机器学习方法在实际应用中受到限制。基于贝叶斯理论的高斯过程回归(Gaussian process regression,GPR),因可输出估计结果的不确定...锂电池健康状态(state of health, SOH)的退化过程在一定程度上是一个非平稳随机过程,使得当前多数点估计机器学习方法在实际应用中受到限制。基于贝叶斯理论的高斯过程回归(Gaussian process regression,GPR),因可输出估计结果的不确定性,近年来在锂电池SOH区间估计中得到广泛应用。然而,GPR的性能很大程度上取决于其核函数的选择,当前研究多凭借经验选用固定单一核函数,无法适应不同的数据集。为此,本文提出一种基于自适应最优组合核函数GPR的锂电池SOH区间估计方法。该方法首先从电池充放电数据中提取出多个健康因子(health factor, HF),并采用皮尔森相关系数法优选出6个与SOH高度相关的健康因子作为模型的输入。然后,在当前常用的7个核函数集合上,通过两两随机组合构造新的组合核函数,并利用交叉验证自适应优选出最优组合核函数。采用3个不同数据集对所提方法进行了验证,结果表明:本文方法具有出色的SOH区间估计性能。在3个公开数据集上,平均区间宽度指标在0.0509以内,平均区间分数大于-0.0004,均方根误差小于0.0181。展开更多
基金supported by the Natural Science Foundation of Ningbo,China (Grant Nos.2009A610014 and 2009A610154)the Natural Science Foundation of Zhejiang Province,China (Grant No.Y6090131)
文摘Steady-state heat conduction problems arisen in connection with various physical and engineering problems where the functions satisfy a given partial differential equation and particular boundary conditions, have attracted much attention and research recently. These problems are independent of time and involve only space coordinates, as in Poisson＇s equation or the Laplace equation with Dirichlet, Neuman, or mixed conditions. When the problems are too complex, it is difficult to find an analytical solution, the only choice left is an approximate numerical solution. This paper deals with the numerical solution of three-dimensional steady-state heat conduction problems using the meshless reproducing kernel particle method （RKPM）. A variational method is used to obtain the discrete equations. The essential boundary conditions are enforced by the penalty method. The effectiveness of RKPM for three-dimensional steady-state heat conduction problems is investigated by two numerical examples.
文摘锂电池健康状态(state of health, SOH)的退化过程在一定程度上是一个非平稳随机过程,使得当前多数点估计机器学习方法在实际应用中受到限制。基于贝叶斯理论的高斯过程回归(Gaussian process regression,GPR),因可输出估计结果的不确定性,近年来在锂电池SOH区间估计中得到广泛应用。然而,GPR的性能很大程度上取决于其核函数的选择,当前研究多凭借经验选用固定单一核函数,无法适应不同的数据集。为此,本文提出一种基于自适应最优组合核函数GPR的锂电池SOH区间估计方法。该方法首先从电池充放电数据中提取出多个健康因子(health factor, HF),并采用皮尔森相关系数法优选出6个与SOH高度相关的健康因子作为模型的输入。然后,在当前常用的7个核函数集合上,通过两两随机组合构造新的组合核函数,并利用交叉验证自适应优选出最优组合核函数。采用3个不同数据集对所提方法进行了验证,结果表明:本文方法具有出色的SOH区间估计性能。在3个公开数据集上,平均区间宽度指标在0.0509以内,平均区间分数大于-0.0004,均方根误差小于0.0181。
