基于混合序列模型的电动汽车续驶里程预测方法及系统
By employing a prediction method based on a hybrid sequence model, combining LSTM, Transformer, and a parameterless attention module, the problem of long-term dependencies in electric vehicle range prediction is solved, achieving higher accuracy in range prediction and improving the range prediction capability of electric vehicles.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CATARC AUTOMOTIVE TEST CENT TIANJIN CO LTD
- Filing Date
- 2026-05-06
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies suffer from insufficient prediction accuracy in predicting the driving range of electric vehicles, especially when dealing with complex multi-factor interactions. Traditional models such as RNNs have limitations in handling long-term dependencies.
A prediction method based on a hybrid sequence model is adopted, which combines LSTM, Transformer and parameterless attention module. Through hierarchical feature extraction and cross-scale feature fusion, a prediction model with multi-branch parallel output is constructed. Considering multi-source data and temperature correction, the prediction accuracy is improved.
It significantly improves the accuracy of driving range prediction, reducing the mean absolute error and root mean square error to 3.5 km and 4.2 km, respectively. The prediction accuracy reaches 89.5% under complex road conditions and the error drops to 4.1 km under extreme temperature conditions. The overall prediction accuracy and stability are better than the traditional RNN model.
Smart Images
Figure CN122135458B_ABST