摘要
准确的电池荷电状态(SOC)估计是电动车辆正常工作的基本前提。针对目前电池荷电状态估计时存在的非线性、不平稳等干扰因素的影响,本工作提出了基于灰狼优化算法的极限学习机的锂离子电池SOC估计方法,以提高估计精度并缩短估计时长。传统的极限学习机(ELM)直接随机生成模型参数,并对SOC进行估计,该方法运行速度快且泛化性能好。但极限学习机需要找出最优的隐含层神经元参数才能达到较高的精度。因此,通过灰狼优化算法(GWO)进一步优化模型参数,并通过选择合适的激活函数,弥补了传统极限学习机的不足。最后通过与粒子群优化的前馈神经网络算法(BPNN-PSO)和极限学习机算法从多维度进行对比分析,在不同工况下验证了此方法在电池SOC估计中的优越性。结果表明,基于灰狼优化算法的极限学习机的锂离子电池荷电状态估计精度高、估算时长较短且鲁棒性较好,明显优于传统SOC估计方法。本研究有助于推动新能源车辆电池管理系统的开发与应用,为可靠的电池管理系统的研发提供支持。
Accurate estimation of battery state of charge(SOC)is the basic premise for the normal operation of electric vehicles.An extreme learning machine(ELM)based on the gray wolf optimizer(GWO)algorithm was proposed to estimate the SOC of lithium-ion batteries,improve the estimation accuracy,and shorten the estimation time.The traditional ELM generates model parameters randomly.This method runs fast and has good generalization performance.However,the ELM must determine the optimal hidden layer neuron parameters to achieve high accuracy.Therefore,the GWO algorithm was adopted to further optimize the model parameters.The appropriate activation function was selected to make up for the shortcomings of the traditional ELM.Finally,the advantages of this method in battery SOC estimation are verified under different working conditions by comparison with the particle swarm optimization-feedforward neural network algorithm(BPNN-PSO)and the ELM.The result shows that the ELM based on the GWO algorithm has high accuracy,short estimation time,and good robustness,which is better than those of the traditional SOC estimation method.This research helps promote the development and application of new energy vehicle battery management systems and provide support for the research and development of reliable battery management systems.
作者
王桥
魏孟
叶敏
李嘉波
徐信芯
WANG Qiao;WEI Meng;YE Min;LI Jiabo;XU Xinxin(National Engineering Laboratory for Highway Maintenance Equipment,Chang'an University,School of Construction Machinery,Xi'an 710064,Shaanxi,China)
出处
《储能科学与技术》
CAS
CSCD
北大核心
2021年第2期744-751,共8页
Energy Storage Science and Technology
基金
国家自然科学基金青年项目(51805041)
河南省重大科技专项(191110211500)
长安大学研究生科研创新实践项(300103703038)。
关键词
锂离子电池
荷电状态
极限学习机
灰狼优化
lithium-ion battery
state of charge
extreme learning machine
gray wolf optimizer
