【摘要】 电动汽车续驶里程预测是驾驶者最关心的问题之一，为解决现有预测算法模型精度低、相对误差大，无法有效缓解用户“里程焦虑”的问题。提出了基于改进GRU的电动汽车续驶里程智能预测方法，该模型引入双向记忆改进策略、多尺度深度特征提取策略和多头外部注意力机制改进策略，实现续驶里程精准预测的同时保证预测实时性。消融实验表明，3个改进策略均能有效提升预测准确度，共同作用时，MSE误差较改进前降低72.2%;对比实验表明，该算法的MAE误差仅为9.82,优于主流深度学习预测算法，且预测速度能够满足实时性需求。该算法进一步实现电动汽车续驶里程的准确预测，为电动汽车的信息化管控和智能化建设提供指导意见，具有突出的工程应用意义。更多还原

【Abstract】 Driving range prediction of electric vehicles is one of the most concerned problems for drivers.To solve the problem of "range anxiety",which have not been effectively fixed, we propose an intelligent method based on improved GRU for driving range prediction.This model introduces bi-directional memory improvement strategy, multi-scale feature extraction strategy and multi-head external attention mechanism improvement strategy to achieve accurate prediction of driving range and ensure real-time performance.Ablation experiments show that each strategy can effectively improve the prediction accuracy.When the three work together, the MSE is reduced by 72.2% compared with GRU.Comparative experiments show that this algorithm, whose MAE is only 9.82,outperforms the mainstream prediction algorithm and the prediction speed can meet the real-time demand.This method can further achieve the accurate driving range prediction and provide guidance for the information management and intelligent construction of electric vehicles, having outstanding engineering application significance.更多还原