# Feature parameters Unit Data Source
1 Average time consumption (sec//km) VIL
2 Average acceleration (m//s2) VIL
3 Average deceleration (m//s2) VIL
4 Proportion of idling stop (%) VIL
5 Proportion in velocity 0-20km//h (%) VIL
6 Proportion in velocity 20-40km//h (%) VIL
7 Proportion in velocity 40-60km//h (%) VIL
8 Proportion in velocity > 60km//h (%) VIL
9 Average gas pedal position (%) VIL
10 Average energy consumption (kJ//km) VIL
11 Traffic flow status (-) Baidu Map
12 Traffic jam direction (-) Baidu Map
13 Traffic flow speed (km//h) Baidu Map| $\#$ | Feature parameters | Unit | Data Source |
| :---: | :--- | :--- | :---: |
| 1 | Average time consumption | $(\mathrm{sec} / \mathrm{km})$ | VIL |
| 2 | Average acceleration | $(\mathrm{m} / \mathrm{s} 2)$ | VIL |
| 3 | Average deceleration | $(\mathrm{m} / \mathrm{s} 2)$ | VIL |
| 4 | Proportion of idling stop | $(\%)$ | VIL |
| 5 | Proportion in velocity $0-20 \mathrm{~km} / \mathrm{h}$ | $(\%)$ | VIL |
| 6 | Proportion in velocity $20-40 \mathrm{~km} / \mathrm{h}$ | $(\%)$ | VIL |
| 7 | Proportion in velocity $40-60 \mathrm{~km} / \mathrm{h}$ | $(\%)$ | VIL |
| 8 | Proportion in velocity $>60 \mathrm{~km} / \mathrm{h}$ | $(\%)$ | VIL |
| 9 | Average gas pedal position | $(\%)$ | VIL |
| 10 | Average energy consumption | $(\mathrm{kJ} / \mathrm{km})$ | VIL |
| 11 | Traffic flow status | $(-)$ | Baidu Map |
| 12 | Traffic jam direction | $(-)$ | Baidu Map |
| 13 | Traffic flow speed | $(\mathrm{km} / \mathrm{h})$ | Baidu Map |
图 10.子条件的能量/时间消耗分布。
函数为类似高斯的径向基函数,初始中心参数为随机选择。在 RBF NN 的训练过程中,表 I 中的交通流信息(#11~#13)是输入,而输出则是聚类样本数据图 10 中对应的子条件索引。经过训练后,RBF NN 可以利用实时交通流信息( Stat_("Trff "),Drc_(Jam)\operatorname{Stat}_{\text {Trff }}, D r c_{\mathrm{Jam}} 和 Spd_("Trff ")S p d_{\text {Trff }} )来预测整个路线(长期)的未来行车条件子类型( Sub_("Prd ")S u b_{\text {Prd }} )。根据图 10 中不同子条件的概率分布,可以预测长期的能量和时间消耗。以图 11 为例,当车辆到达不同路线段时,会触发三次长期预测,紫色竖线表示长期预测。由于使用了图 10 中的概率分布,时间和能耗预测结果呈黄色带状。在图 11 中,实际值在黄带范围内,接近平均值,证明了长期预测的准确性。此外,在执行新的预测时,累积的预测误差会被消除,这使得黄色跨度突然缩小为零。
成本函数: Pr_(H2)xx fuel+Pr_(fc)xx Delta SOH_(fc)+Pr_("bat ")xx Delta SOH_("bat ")P r_{\mathrm{H} 2} \times f u e l+P r_{\mathrm{fc}} \times \Delta S O H_{\mathrm{fc}}+P r_{\text {bat }} \times \Delta S O H_{\text {bat }}
Pr_(H2),Pr_("fc "),Pr_("bat ")P r_{\mathrm{H} 2}, P r_{\text {fc }}, P r_{\text {bat }}
[SOC_("lb "),SOC_("ub ")]\left[S O C_{\text {lb }}, S O C_{\text {ub }}\right]
参考 SOC 的上下限 ( SOC_("ref ")S O C_{\text {ref }} )
Symbol Definition and meaning
n Stage index: defined in Fig. 1, where 1 <= n <= N
N Num of stage: refer to num of route segments in Fig. 1
s_(n) System state: refer to battery SOC
M Num of system state: feasible state within [State l_(lb), State _(ub) ]
[State e_("lb "), State {:_("ub ")] Feasible domain of s_(n) during SDP calculation
f_(n)(s_(n),x_(n)) Optimal solution with decision x_(n):f_(n)(s_(n),x_(n))=sum_(i=1)^(S)p_(i)(C_(i)+f_(n+1)^(**)(i):} )
x_(n) Decision variable: refer to the feasible charging options
D The size of the policy decision: D=51
p_(i) Probability: refer to probability distribution in Fig. 8
S Num of probability: refer to the num of intervals in Fig. 8
C_(i) Costs function: Pr_(H2)xx fuel+Pr_(fc)xx Delta SOH_(fc)+Pr_("bat ")xx Delta SOH_("bat ")
Pr_(H2),Pr_("fc "),Pr_("bat ") Monetary prices of H_(2) fuel, fuel cell stack, and power battery
f_(n+1)^(**)(i) Optimal solution from stage n+1 to end: f_(n+1)^(**)(i)=min_(x_(n+1))f_(n+1)(i,x_(n+1))
[SOC_("lb "),SOC_("ub ")] Upper and lower boundary of the reference SOC ( SOC_("ref ") )| Symbol | Definition and meaning |
| :---: | :---: |
| $n$ | Stage index: defined in Fig. 1, where $1 \leq \mathrm{n} \leq \mathrm{N}$ |
| $N$ | Num of stage: refer to num of route segments in Fig. 1 |
| $s_{n}$ | System state: refer to battery SOC |
| M | Num of system state: feasible state within [State $\mathrm{l}_{\mathrm{lb}}$, State $_{\mathrm{ub}}$ ] |
| [State $e_{\text {lb }}$, State $\left._{\text {ub }}\right]$ | Feasible domain of $s_{n}$ during SDP calculation |
| $f_{n}\left(s_{n}, x_{n}\right)$ | Optimal solution with decision $x_{\mathrm{n}}: f_{n}\left(s_{n}, x_{n}\right)=\sum_{i=1}^{S} p_{i}\left(C_{i}+f_{n+1}^{*}(\mathrm{i})\right.$ ) |
| $x_{n}$ | Decision variable: refer to the feasible charging options |
| D | The size of the policy decision: $D=51$ |
| $p_{i}$ | Probability: refer to probability distribution in Fig. 8 |
| $S$ | Num of probability: refer to the num of intervals in Fig. 8 |
| $C_{i}$ | Costs function: $P r_{\mathrm{H} 2} \times f u e l+P r_{\mathrm{fc}} \times \Delta S O H_{\mathrm{fc}}+P r_{\text {bat }} \times \Delta S O H_{\text {bat }}$ |
| $P r_{\mathrm{H} 2}, P r_{\text {fc }}, P r_{\text {bat }}$ | Monetary prices of $\mathrm{H}_{2}$ fuel, fuel cell stack, and power battery |
| $f_{n+1}^{*}(i)$ | Optimal solution from stage $n+1$ to end: $f_{n+1}^{*}(i)=\min _{x_{n+1}} f_{n+1}\left(i, x_{n+1}\right)$ |
| $\left[S O C_{\text {lb }}, S O C_{\text {ub }}\right]$ | Upper and lower boundary of the reference SOC ( $S O C_{\text {ref }}$ ) |
大小 NN 计算总循环操作。根据 [38],SDP 的总循环等于 S*D*M*NS \cdot D \cdot M \cdot N 。
如图 12 右侧所示,状态大小 MM 是指状态 _(lb)_{\mathrm{lb}} 和状态 _(ub)_{\mathrm{ub}} 之间黄色带的高度。因此, MM 可以看作是 NN 的片断线性函数:
M(N)={[kN,N in(I)],[kN//2],[" constant "," when "],[N in(" II ")],[N in(III)]:}M(N)= \begin{cases}k N & N \in(\mathrm{I}) \\ k N / 2 \\ \text { constant } & \text { when } \\ N \in(\text { II }) \\ N \in(\mathrm{III})\end{cases}
结合公式(4)和 O(M*N)O(M \cdot N) ,本研究中 SDP 的复杂度应为
Complexity(N)={[O(N^(2)),N in(I)],[O(N^(2)//2)," when ",N in(II)],[O(N),N in(III)]:}\operatorname{Complexity}(N)=\left\{\begin{array}{lll}
O\left(N^{2}\right) & N \in(\mathrm{I}) \\
O\left(N^{2} / 2\right) & \text { when } & N \in(\mathrm{II}) \\
O(N) & N \in(\mathrm{III})
\end{array}\right.
其中 NN 为问题规模,复杂度分为三个部分。这是因为系统状态 s_(n)s_{n} 指的是电池 SOC,其最大可行区域限制为
[0.1,0.9],状态大小 MM 不能随着阶段大小 NN 的扩大而不断扩大。如图 12 所示(假设 SOC_("end ")=0.3S O C_{\text {end }}=0.3 ),当 n=7n=7 时,状态的下边界状态 _(lb)_{\mathrm{lb}} 达到极限,当 n=46n=46 时,状态的上边界状态 _(ub)_{\mathrm{ub}} 达到极限。
T_(cmpt)(N)=T_(O1)xxsum_(j=1)^(N)(10^(3)xx(" State "_(ub)(j)-" State "_(lb)(j))+1)T_{\mathrm{cmpt}}(N)=T_{\mathrm{O} 1} \times \sum_{j=1}^{N}\left(10^{3} \times\left(\text { State }_{\mathrm{ub}}(j)-\text { State }_{\mathrm{lb}}(j)\right)+1\right)
{:[C(D_(chg))=SOC_(Err)(D_(chg))xxw_(SOC)+fuel(D_(chg))xx Pr_(fuel)],[+Delta SOH_(fc)(D_(chg))xx Pr_(fc)+Delta SOH_(bat)(D_(chg))xx Pr_(bat)]:}\begin{aligned}
C\left(D_{\mathrm{chg}}\right)= & S O C_{\mathrm{Err}}\left(D_{\mathrm{chg}}\right) \times w_{\mathrm{SOC}}+\operatorname{fuel}\left(D_{\mathrm{chg}}\right) \times P r_{\mathrm{fuel}} \\
& +\Delta S O H_{\mathrm{fc}}\left(D_{\mathrm{chg}}\right) \times P r_{\mathrm{fc}}+\Delta S O H_{\mathrm{bat}}\left(D_{\mathrm{chg}}\right) \times P r_{\mathrm{bat}}
\end{aligned}
其中
{:[SOC_(Err)(D_(chg))],[={[SOC_(lb)-SO^(˙)C(D_(chg))","," if "SO^(˙)C(D_(chg)) <= SOC_(lb)],[0","," if "SOC_(lb) <= SO^(˙)C(D_(chg)) <= SOC_(ub)],[SO^(˙)C(D_(chg))-SOC_(ub)","," if "SO^(˙)C(D_(chg)) >= SOC_(ub)]:}]:}\begin{aligned}
& S O C_{\mathrm{Err}}\left(D_{\mathrm{chg}}\right) \\
& = \begin{cases}S O C_{\mathrm{lb}}-S \dot{O} C\left(D_{\mathrm{chg}}\right), & \text { if } S \dot{O} C\left(D_{\mathrm{chg}}\right) \leq S O C_{\mathrm{lb}} \\
0, & \text { if } S O C_{\mathrm{lb}} \leq S \dot{O} C\left(D_{\mathrm{chg}}\right) \leq S O C_{\mathrm{ub}} \\
S \dot{O} C\left(D_{\mathrm{chg}}\right)-S O C_{\mathrm{ub}}, & \text { if } S \dot{O} C\left(D_{\mathrm{chg}}\right) \geq S O C_{\mathrm{ub}}\end{cases}
\end{aligned}
其次,成本函数 C(D_("chg "))C\left(D_{\text {chg }}\right) 的定义见 (8),其中包含四个因子。后三个因子的定义与 Talbe II 中的顶层 EMS 相同。对于第一个因素, SOC_("Err ")S O C_{\text {Err }} 是来自 SOC_("ref "),w_("SOC ")S O C_{\text {ref }}, w_{\text {SOC }} 的误差, SO^(˙)CS \dot{O} C 是当前路线段结束时的电池 SOC 预测值。最小化 (8) 可确保实时 SOC 在一般情况下遵循 SOC_("ref ")S O C_{\text {ref }} 。
P^(˙)_(bat)(k)={[P^(˙)_(mot)(k)//eta_(mot)-P_(fc)^(**)","," if in "D_(chg)],[P^(˙)_(mot)(k)//eta_(mot)","," if not "]:}\dot{P}_{\mathrm{bat}}(k)= \begin{cases}\dot{P}_{\mathrm{mot}}(k) / \eta_{\mathrm{mot}}-P_{\mathrm{fc}}^{*}, & \text { if in } D_{\mathrm{chg}} \\ \dot{P}_{\mathrm{mot}}(k) / \eta_{\mathrm{mot}}, & \text { if not }\end{cases}
[1] D.-D.Tran, M. Vafaeipour, M. El Baghdadi, R. Barrero, J. Van Mierlo, and O. Hegazy, "Thorough state-of-the-art analysis of electric and hybrid vehicle powertrains:Topologies and integrated energy management strategies," Renew.Sustain.能源评论》,第 119 卷,2020 年 3 月,艺术编号 109596。
[2] H. Dong, J. Fu, Z. Zhao, Q. Liu, Y. Li, and J. Liu, "A comparative study on the energy flow of a conventional gasoline-powered vehicle and a new dual clutch parallel-series plug-in hybrid electric vehicle under NEDC," Energy Convers.管理》,第 218 卷,2020 年 8 月,艺术编号 113019。
[3] P. Saiteja 和 B. Ashok,"关于混合动力汽车结构架构、能源控制策略和优化能源管理开发过程的重要评论",Renew.Sustain.能源评论》,第 157 卷,2022 年 4 月,艺术编号 112038。
[4] B. Hu, S. Zhang, and B. Liu, "A hybrid algorithm combining data-driven and simulation-based reinforcement learning approaches to energy management of hybrid electric vehicles," IEEE Trans.Transport.Electrific.》,第 10 卷,第 1 期,第 1257-1273 页,2024 年 3 月。
[5] Y. Zhou、A. Ravey 和 M.-C.Péra, "A survey on driving prediction techniques for predictive energy management of plug-in hybrid electric vehicles," J. Power Sources, vol. 412, pp.
[6] Z. Yao 等人,"基于交错式三电平 DC-DC 转换器中近 CRM 的电压和电流无传感器同步自平衡机制",《IEEE Trans.39卷,第2期,第2086-2099页,2024年2月。
[7] T. Teng, X. Zhang, H. Dong, and Q. Xue, "A comprehensive review of energy management optimization strategies for fuel cell passenger vehicle," Int. J. Hydrogen Energy, vol. 45, no.J. Hydrogen Energy, vol. 45, no.39, pp.
[8] P. Dong 等人,"基于智能互联技术的混合动力电动汽车能源管理策略的实际应用:Development stages, challenges, and future trends," Renew.Sustain.能源评论》,第 170 卷,2022 年 12 月,艺术编号 112947。
[9] F. Zhang, X. Hu, R. Langari, and D. Cao, "Energy management strategies of connected HEVs and PHEVs:最新进展与展望》,《能源燃烧科学进展》,第 73 卷,第 1 页。能源燃烧科学》,第 73 卷,第 235-256 页,2019 年 7 月。
[10] W. Li, F. Qiu, L. Li, Y. Zhang, and K. Wang, "Simulation of vehicle interaction behavior in merging scenarios:结合博弈论的深度最大熵反强化学习方法",IEEE Trans.Intell.Vehicles, vol. 9, no. 1, pp.
[11] W. Li, Y. Zhang, L. Li, Y. Lv, and M. Wang, "A pedestrian trajectory prediction model for right-turn unsignalized intersections based on game theory," IEEE Trans.Intell.Transp.Syst.》,第 25 卷第 8 期,第 9643-9658 页,2024 年 8 月。
[12] S. Xie, S. Qi, K. Lang, X. Tang, and X. Lin, "Coordinated management of connected plug-in hybrid electric buses for energy saving, intervehicle safety, and battery health," Appl.
[13] W. Cui, N. Cui, T. Li, Z. Cui, Y. Du, and C. Zhang, "An efficient multiobjective hierarchical energy management strategy for plug-in hybrid electric vehicle in connected scenario," Energy, vol. 257, Oct. 2022, Art. No.
[14] L. Yang, W. Wang, C. Yang, X. Du, and W. Zhang, "Online mixed-integer optimal energy management strategy for connected hybrid electric vehicles," J. Cleaner Prod., vol. 374, Nov. 2022, Art. No.
[15] R. Tian、C. Wang、J. Hu 和 Z. Ma,"用于交通预测的多尺度时空感知变换器",《信息科学》,第 648 卷,2023 年 11 月,艺术编号 119557。科学》,第 648 卷,2023 年 11 月,艺术编号 119557。
[20] J. Hou and Z. Song,"A hierarchical energy management strategy for hybrid energy storage via vehicle-to-cloud connectivity",Appl. Energy,vol. 257,Jan. 2020,Art. no.113900。
[21] W. Cui 等:《网络物理交互的互联插电式混合动力电动汽车多目标分层能源管理》,《应用能源》,第 360 卷,第 122816-122833 页,2024 年 4 月。
[22] J. Shi, B. Xu, X. Zhou, and J. Hou, "A cloud-based energy management strategy for hybrid electric city bus considering real-time passenger load prediction," J. Energy Storage, vol. 45, Jan. 2022, Art. No.
[23] Y. Yu, J. Jiang, H. Min, Z. Zhang, W. Sun, and Q. Cao, "Integrated energy and thermal management strategy for extended range electric vehicle based on battery temperature and state-of-charge global planning," Energy Convers.管理》,第 288 卷,2023 年 7 月,艺术编号 117154。
[24] Y. Jin 等人,"基于全局最优能耗和充电状态轨迹预测的互联插电式混合动力电动汽车的双尺度能源管理",《能源》,第 267 卷,2023 年 3 月,艺术品编号:126498。
[26] Y. Zhang, S. Lu, Y. Yang, and Q. Guo, "Internet-distributed vehicle-in-the-loop simulation for HEVs," IEEE Trans.Veh. Technol.Veh. Technol.5,第 3729-3739 页,2018 年 5 月。
[27] R. Feng 等人,"混合动力重型卡车 AMT 和 MT 换挡模式下本地行驶测试循环能量流比较研究",《能源管理》,第 256 卷,2022 年 3 月,艺术编号 115359。管理》,第 256 卷,2022 年 3 月,艺术编号:115359。
[29] Y. Zhang, H. Liu, Z. Zhang, Y. Luo, Q. Guo, and S. Liao, "Cloud computing-based real-time global optimization of battery aging and energy consumption for plug-in hybrid electric vehicles," J. Power Sources, vol. 479, Dec. 2020, Art.
[30] D. Andre、C. Appel、T. Soczka-Guth 和 D. U. Sauer,《锂离子电池 SOC 和 SOH 估算的高级数学方法》,《电源》杂志,第 224 卷,第 20-27 页,2013 年 2 月。
[31] W. Qi 等人,"用于动力电池热管理的液冷板中的多 U 型微通道",《应用热工》,第 256 卷,2024 年 11 月,艺术编号 123984,doi:10.1016/j. applthermaleng.2024.123984。
[32] P. Pei, Q. Chang, and T. Tang, "A quick evaluating method for automotive fuel cell lifetime," Int.33, no. 14, pp.
[33] P. C. Pei 等:《质子交换膜燃料电池堆在车辆应用中的寿命评估方法》,中国专利 GB/T 389142 020,2020 年 12 月 1 日。
[34] R. Shankar 和 J. Marco,"真实驾驶条件下电动汽车和插电式混合动力电动汽车能耗估算方法",IET Intell.Transp.Syst.》,第 7 卷,第 1 期,第 138-150 页,2013 年 3 月。
[36] N. Yang, L. Han, C. Xiang, H. Liu, T. Ma, and S. Ruan, "Real-time energy management for a hybrid electric vehicle based on heuristic search," IEEE Trans.Veh. Technol.71, no. 12, pp.
[37] F. Schwenker、H. A. Kestler 和 G. Palm,"径向基函数网络的三个学习阶段",《神经网络》,第 14 卷,第 4-5 期,第 439-458 页,2001 年 5 月。
[38] T. H. Cormen 等,《算法导论》,第三版,美国马萨诸塞州剑桥:麻省理工学院出版社,2022 年。
[39] C. Zhao, X. Shu, X. Yan, X. Zuo, and F. Zhu, "RDD-YOLO: A modified YOLO for detection of steel surface defects," Measurement, vol. 214, Jun. 2023, Art. No.
[41] H. Kawai、Z. Zhang、R. Kennel 和 S. Doki,"基于有限控制集模型预测控制与电压平滑器的直接速度控制",IEEE Trans.Ind.Electron.3,第 2363-2372 页,2022 年 5 月。
[42] P. Pisu 和 G. Rizzoni,"混合动力电动汽车监督控制策略的比较研究",IEEE Trans.Control Syst.技术》,第 15 卷,第 3 期,第 506-518 页。3,第 506-518 页,2007 年 5 月。
[43] Z. Nie, Y. Jia, W. Wang, and R. Outbib, "Eco-Co-Optimization strategy for connected and automated fuel cell hybrid vehicles in dynamic urban traffic settings," Energy Convers.管理》,第 263 卷,2022 年 7 月,艺术编号 115690。