| Hierarchical Control for PHEV Platoon Based on Multi-information Fusion Speed Prediction |
| Yanli Yin1,2, Haixin Chen1, Fuchun Zhang1, Fuzhen Wang1, Hangyang Xiao1 |
1School of Mechatronics and Automobile Engineering, Chongqing Jiao Tong University, Chongqing 400054, China
2Key Laboratory of Advanced Manufacturing Technology of the Ministry of Education, Guizhou University, Guiyang 550025, China |
|
|
|
Received: September 24, 2024; Revised: October 25, 2024 Accepted: October 30, 2024. Published online: January 7, 2025. |
|
|
|
| ABSTRACT |
|
To address the issues of speed planning and driving condition adaptability in the platoon operation of plug-in hybrid electric vehicles (PHEVs), a hierarchical control strategy for platooning based on multi-information-fused speed prediction is proposed. First, historical time series data from VISSIM is used to predict the speed of the lead vehicle in the platoon with a Long Short-Term Memory (LSTM) neural network. Second, the upper controller employs vehicle-to-vehicle (V2V) communication to obtain information about preceding vehicles and determines the optimal following speed using the Linear Model Predictive Control (LMPC) algorithm. Then, the lower controller applies Principal Component Analysis (PCA) and K-means clustering to establish four typical operating conditions, training an offline Deep Q-Learning (DQL) network and condition recognition model for each, enabling adaptive energy management. Finally, the results show that the LSTM model improves prediction accuracy by 28.51% over the Recurrent Neural Network (RNN). Compared to the baseline CD-CS strategy, the DQL strategy achieves a 20.31% reduction in fuel consumption per 100 km, while the proposed adaptive condition recognition-based energy management strategy (A-DQL) reduces fuel consumption by 22.51%, demonstrating improved fuel economy. |
| Key Words:
Vehicle platoon · VISSIM simulation · Speed prediction · Driving condition identification · Model predictive control (MPC) · DQL algorithm |
|
Corresponding Author.
Yanli Yin , Email. 
Preview 
Full text via DOI 
Download Citation
Print
