基于DE-ELM的电池SOC预测研究  

The Prediction of SOC of Battery Based on DE-ELM

作　　者：罗宏远 王德运[1,2] 潘雯雯[1,2] 魏帅[1,2] LUO Hong-yuan WANG De-yun PAN Wen-wen WEI Shuai(School of Economics and Management Science, China University of Geosciences, Wuhan 430074, China Mineral Resource Strategy and Policy Research Center of China University of Geosciences, Wuhan 430074 China)

机构地区：[1]中国地质大学(武汉)经济管理学院,湖北武汉430074 [2]中国地质大学(武汉)中国矿产资源战略与政策研究中心,湖北武汉430074

出　　处：《数学的实践与认识》2016年第22期236-243,共8页Mathematics in Practice and Theory

基　　金：国家自然科学基金(71301153);中央高校基本科研业务费专项资金(CUG140612)

摘　　要：针对极限学习机的随机性较大的问题,提出一种基于差分演化的极限学习机算法模型(DE-ELM).采用差分演化算法(DE)对极限学习机(ELM)随机给定的输入权值矩阵和隐含层阈值进行寻优,降低了随机性给ELM造成的影响,减少ELM网络震荡,提高了ELM预测精度.并且将DE-ELM应用在电池SOC的预测上,同时与ELM和BP神经网络的预测进行了对比,结果表明:DE-ELM在电池SOC预测上的表现优于ELM和BP神经网络,能满足电池SOC的预测精度要求.In view of the problem of the randomness of the extreme learning machine,a model of extreme learning machine based on differential evolution（DE-ELM） is proposed in this paper.After the input weight matrix and hidden layer threshold of ELM are optimized by DE,the effects of its randomness of ELM were reduced and the prediction accuracy of the ELM is improved,we use DE-ELM to estimate the SOC of battery,and compare to the results of ELM and BP- Neural Network.The result shows that：DE-ELM is better than ELM and BP- Neural Network,and can satisfy the requirement of the estimation for SOC of battery.

关 键 词：极限学习机 差分演化 SOC预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]