摘要
针对锂电池荷电状态(SOC)无法直接测量的问题,提出一种使用改进蝗虫优化算法(IGOA)优化轻量级梯度提升机器学习(LightGBM)的SOC预测模型。首先,引入非线性递减系数、自适应权重系数和位置更新策略改进标准GOA算法,获取最优种群个体。其次,利用IGOA寻找LightGBM算法中的最优超参数组合,建立IGOA-LightGBM预测模型;最后,在马里兰大学提供的电池数据集上,将IGOA-LightGBM、LightGBM和BP神经网络模型的预测结果进行对比。实验结果表明,IGOA-LightGBM模型表现最优,平均绝对误差、平均绝对值百分比误差和均方根误差分别为0.013%、0.022%、0.018%,具备良好的预测精度及工程研究意义。
In response to the problem of being unable to directly measure the state of charge(SOC)of lithium batteries,a method is proposed to use the Improved Grasshopper Optimization Algorithm(IGOA)to optimize the Light Gradient Boosting Machine for SOC prediction.Firstly,a nonlinear decreasing coefficient,adaptive weight coefficient,and position update strategy are introduced to improve the standard IGOA to obtain the optimal population individuals.Secondly,the IGOA is used to find the optimal hyperparameter combination in the LightGBM algorithm to establish the IGOA-LightGBM prediction model.Finally,the prediction results of the IGOA-LightGBM,LightGBM,and BP neural network models are compared,using a battery dataset provided by the University of Maryland.Experimental results show that the IGOA-LightGBM model performs the best,with average absolute error,average absolute percentage error,and root mean square error of 0.013%,0.022%,and 0.018%respectively,demonstrating good prediction accuracy and engineering research significance.
作者
任小强
何青
聂清彬
REN Xiao-qiang;HE Qing;NIE Qing-bin(Hope College,Southwest Jiaotong University,Chengdu 610400,China)
出处
《广州城市职业学院学报》
2024年第1期91-95,共5页
Journal Of Guangzhou City Polytechnic
基金
西南交通大学希望学院青年科研课题“基于充电需求预测的电动汽车充电站选址规划”(编号:2023050)
西南交通大学希望学院质量工程课题“产教融合背景下的‘机器学习’课程改革与实践”(编号:2023036)
中国电子劳动学会课题“校企合作背景下的‘机器学习’课程改革与实践”(编号:Cea12023021)
成都市哲学社会科学重点研究基地成都市交通+旅游大数据应用技术研究课题“基于多源异构大数据平台的美丽乡村活力指数预测方法研究”(编号:20231012)。
关键词
锂离子电池
荷电状态
蝗虫优化算法
轻量级梯度提升机
lithium ion battery
state of charge
Improved Grasshopper Optimization Algorithm
Lightweight Gradient Boosting Machine
