摘要
锂离子电池组的荷电状态(SOC)估算精度是制约电动汽车发展的关键技术之一,它影响汽车的行驶安全、电池寿命和均衡器效果等关键性能。设计了一种主元分析法(PCA)-蚁群算法(ACO)-Elman动态神经网络-无迹卡尔曼滤波(UKF)算法。试验结果表明,通过算法间取长补短,既可以克服系统输入变量维数大、Elman神经网络易陷入局部最优、UKF过于依赖模型等缺陷,又使均方差维持在1%之内,提高了算法的稳定性,且复杂程度较低。通过Advisor 2002软件验证结果表明,在实际行驶工况下,虽然误差曲线波动略大,但后期曲线逐渐收敛,且估算精度达到0.9%,具备良好的泛化能力和鲁棒性,故有一定的实用意义。
The estimation accuracy of the SOC(state of charge)of a lithium-ion battery pack is one of the key technologies restricting the development of electric vehicles,and it affects key performances such as vehicle driving safety,battery life,and equalizer effect.Based on this,a PCA-ACO-Elman dynamic neural network-UKF algorithm was designed.The experimental results show that the shortcomings of the algorithm are to overcome the shortcomings of the system such as the large dimension of the input variables,the Elman neural network being easily trapped in the local optimum,and the UKF(Unscented Kalman Filter)being too dependent on the model,while maintaining the mean square error within 1%,improving the stability of the algorithm,and the complexity is low.Finally,it is verified by Advisor 2002 software that under actual driving conditions,although the error curve fluctuates slightly,the curve gradually converges in the later period,and the estimation accuracy reaches 0.9%.It has good generalization ability and robustness,so it has certain practical significance.
作者
王伯瑞
郑培
WANG Bo-rui;ZHENG Pei(College of Energy and Power Engineering,Inner MongoliaUniversity of Technology,Hohhot Inner Mongolia 010051,China)
出处
《电源技术》
CAS
北大核心
2020年第10期1506-1509,1517,共5页
Chinese Journal of Power Sources
关键词
荷电状态
动态神经网络
无迹卡尔曼滤波
state of charge
dynamic neural network
unscented Kalman filter
