Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network
Abstract Addressing the issues of complex parameter tuning and challenging online application of control algorithms for energy management strategies in current hybrid tractors. Taking high-power diesel-electric hybrid tractor as the research object, dynamic models for the tractor under three operati...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-05-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-02044-5 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849309726826823680 |
|---|---|
| author | Yiwei Wu Yifan Zhao Changkai Wen Junjiang Zhang Xianghai Yan |
| author_facet | Yiwei Wu Yifan Zhao Changkai Wen Junjiang Zhang Xianghai Yan |
| author_sort | Yiwei Wu |
| collection | DOAJ |
| description | Abstract Addressing the issues of complex parameter tuning and challenging online application of control algorithms for energy management strategies in current hybrid tractors. Taking high-power diesel-electric hybrid tractor as the research object, dynamic models for the tractor under three operating conditions—plowing, rotary tillage, and transportation—were established. With engine power as the agent’s control action, drive motor power and battery SOC as state variables, and tractor equivalent fuel consumption and battery SOC as the reward function, an energy management strategy based on deep reinforcement learning was designed using the double deep Q-network (DDQN) algorithm. Finally, a hardware-in-the-loop (HIL) testing platform was set up to conduct HIL tests on the designed tractor model and energy management strategy. The test results demonstrate that, in comparison to the power-following-based energy management strategy, the DDQN strategy: under plowing conditions, achieves a 1.18% savings in SOC, reduces equivalent fuel consumption by 10.40%, and decreases diesel consumption by 4.59%; under rotary tillage conditions, consumes an additional 3.09% of SOC, but reduces equivalent fuel consumption by 9.78% and diesel consumption by 4.77%; and under transportation conditions, consumes an additional 1.49% of SOC, reduces equivalent fuel consumption by 9.57%, and increases diesel consumption by 0.65%. |
| format | Article |
| id | doaj-art-6d853a63f3f64f259ced53a8cb07e2cf |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-6d853a63f3f64f259ced53a8cb07e2cf2025-08-20T03:53:58ZengNature PortfolioScientific Reports2045-23222025-05-0115112210.1038/s41598-025-02044-5Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-networkYiwei Wu0Yifan Zhao1Changkai Wen2Junjiang Zhang3Xianghai Yan4College of Vehicle and Traffic Engineering, Henan University of Science and TechnologyCollege of Vehicle and Traffic Engineering, Henan University of Science and TechnologyCollege of Engineering, China Agricultural UniversityCollege of Vehicle and Traffic Engineering, Henan University of Science and TechnologyCollege of Vehicle and Traffic Engineering, Henan University of Science and TechnologyAbstract Addressing the issues of complex parameter tuning and challenging online application of control algorithms for energy management strategies in current hybrid tractors. Taking high-power diesel-electric hybrid tractor as the research object, dynamic models for the tractor under three operating conditions—plowing, rotary tillage, and transportation—were established. With engine power as the agent’s control action, drive motor power and battery SOC as state variables, and tractor equivalent fuel consumption and battery SOC as the reward function, an energy management strategy based on deep reinforcement learning was designed using the double deep Q-network (DDQN) algorithm. Finally, a hardware-in-the-loop (HIL) testing platform was set up to conduct HIL tests on the designed tractor model and energy management strategy. The test results demonstrate that, in comparison to the power-following-based energy management strategy, the DDQN strategy: under plowing conditions, achieves a 1.18% savings in SOC, reduces equivalent fuel consumption by 10.40%, and decreases diesel consumption by 4.59%; under rotary tillage conditions, consumes an additional 3.09% of SOC, but reduces equivalent fuel consumption by 9.78% and diesel consumption by 4.77%; and under transportation conditions, consumes an additional 1.49% of SOC, reduces equivalent fuel consumption by 9.57%, and increases diesel consumption by 0.65%.https://doi.org/10.1038/s41598-025-02044-5Deep reinforcement learningHybrid tractorEnergy management strategyHardware in the loop |
| spellingShingle | Yiwei Wu Yifan Zhao Changkai Wen Junjiang Zhang Xianghai Yan Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network Scientific Reports Deep reinforcement learning Hybrid tractor Energy management strategy Hardware in the loop |
| title | Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network |
| title_full | Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network |
| title_fullStr | Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network |
| title_full_unstemmed | Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network |
| title_short | Research on energy management strategies for high-power diesel-electric hybrid tractors based on double deep Q-network |
| title_sort | research on energy management strategies for high power diesel electric hybrid tractors based on double deep q network |
| topic | Deep reinforcement learning Hybrid tractor Energy management strategy Hardware in the loop |
| url | https://doi.org/10.1038/s41598-025-02044-5 |
| work_keys_str_mv | AT yiweiwu researchonenergymanagementstrategiesforhighpowerdieselelectrichybridtractorsbasedondoubledeepqnetwork AT yifanzhao researchonenergymanagementstrategiesforhighpowerdieselelectrichybridtractorsbasedondoubledeepqnetwork AT changkaiwen researchonenergymanagementstrategiesforhighpowerdieselelectrichybridtractorsbasedondoubledeepqnetwork AT junjiangzhang researchonenergymanagementstrategiesforhighpowerdieselelectrichybridtractorsbasedondoubledeepqnetwork AT xianghaiyan researchonenergymanagementstrategiesforhighpowerdieselelectrichybridtractorsbasedondoubledeepqnetwork |