基于改进的神经网络算法对电动车续航里程的预测研究  

Research on the Prediction of Electric Vehicle Range Based on an Improved Neural Network Algorithm

作　　者：张国良 张延良 薛雷 李升建 张杨 Guoliang Zhang;Yanlliang Zhang;Lei Xue;Shengjian Li;Yang Zhang(Weichai Power Co.,Ltd.,Electronic Control and Software Research Institute,Weifang,Shandong,261001)

机构地区：[1]潍柴动力股份有限公司电控与软件研究院,山东省潍坊市261001

出　　处：《机械工程与设计（中英文版）》2024年第1期1-6,共6页Mechanical Engineering and Design

摘　　要：电动车的续航里程受电池温度、电流、单体电压、载重、驾驶行为、剩余电量、车速等多个因素的影响,各个因素之间是一种非线性关系。可以利用传统BP神经网络算法对采集数据进行训练得到续航里程训练模型,但BP算法存在局部最优差、收敛速度慢和初始权阀值不易确定的缺点,因此,可以结合遗传算法对BP神经网络的参数进行优化弥补其缺点,同时利用采集的数据对网络进行重复学习,得到更高精度的续航里程训练模型。The driving range of electric vehicles is influenced by various factors such as battery temperature,current,cell voltage,load,driving behavior,remaining battery level,and vehicle speed,and there is a nonlinear relationship between these factors.Traditional Backpropagation(BP)neural network algorithms can be used to train the collected data to obtain a driving range training model.However,the BP algorithm has the drawbacks of being prone to local optima,slow convergence,and difficulty in determining initial weight values.Therefore,it is possible to combine genetic algorithms to optimize the parameters of the BP neural network to compensate for these shortcomings,while also using the collected data to repeatedly train the network to achieve a more accurate driving range training model.

关 键 词：BP算法 遗传算法 续航里程 人工神经网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]