Professional Documents
Culture Documents
1-s2.0-S2352152X23030967-main-2
1-s2.0-S2352152X23030967-main-2
1-s2.0-S2352152X23030967-main-2
Research papers
A R T I C L E I N F O A B S T R A C T
Keywords: Trams with energy storage are popular for their energy efficiency and reduced operational risk. An effective
Energy management strategy energy management strategy is optimized to enable a reasonable distribution of demand power among the
Competitive particle swarm algorithm storage elements, efficient use of energy as well as enhance the service life of the hybrid energy storage system
Tram
(HESS). Thus, an energy management strategy optimization method of HESS based on a fixed power threshold is
Multiple thresholds
developed. Firstly, the energy management strategy optimization models based on single thresholds and multiple
Hybrid energy storage system
thresholds are developed. Then, the improved particle swarm optimization (PSO) algorithm with competing
variable weights is proposed to solve optimization models. Finally, Guangzhou Haizhu tram is used to illustrate
the performance of the developed method, the minimum charge state of the power battery under multiple
thresholds is improved by 23.36 % over that of single threshold, and the total energy consumption of the power
battery pack is reduced by 58.10 %, which shows that the energy management strategy based on multiple
thresholds has better effects in giving full play to the power supply capacity of supercapacitor and reducing the
energy consumption of power battery. Compared with other two methods, the improved PSO method has more
reliable in utilizing more energy from supercapacitor.
1. Introduction supply technology, the vehicle itself is equipped with energy storage
equipment as the power source of the whole vehicle. The characteristics
With the development of energy storage and control technology and of the energy storage equipment of the tram, which is the tram power
their good results in the field of electric vehicles, the technology of supply system, will largely affect the performance of the whole vehicle.
stored energy traction power supply has been greatly developed. Rail Since there is still a lack of a single energy storage element with high
transportation means have been widely used because of their excellent power density and energy density to meet the vehicle operation re
characteristics such as high carrying capacity, high speed, and punctu quirements [6,7]. A common solution for on-board power supply sys
ality. For energy saving and emission reduction in rail transportation, tems is to use two or more energy storage devices in combination to
the development of fuel cell electric locomotives based on renewable synthesize their respective advantages and optimize the energy man
and clean energy, power locomotives using power batteries, and new agement to meet the vehicle's operation requirements. To realize the
locomotives based on various hybrid power have become new technol coordinated work of electric energy storage elements with different
ogies and ways [1,2]. In addition, the use of energy-storage traction characteristics, an effective power management scheme is needed so
power supply technology can simplify the traction power supply facil that the demand power can be reasonably distributed among the energy
ities along the line, reduce the dependence of trains on the power grid, storage elements so that the performance and advantages of the hybrid
adapt to sections where it is difficult to erect power supply contact power system can be more fully utilized [8]. Therefore, the energy
networks, central urban areas and areas with high landscape re management of a hybrid energy storage system (HESS) is a key issue to
quirements, and reduce public safety hazards. Therefore, the use of be studied.
energy-storage traction power supply technology can achieve good re Through the application of effective energy management control
sults in urban construction [3–5]. techniques, the power performance of the HESS is ensured, the power
Tram with energy storage is the application of energy storage power braking energy is effectively utilized and the service life of the HESS is
* Corresponding authors.
E-mail addresses: xqcheng@bjtu.edu.cn (X. Cheng), xingzongyi@163.com (Z. Xing).
https://doi.org/10.1016/j.est.2023.109698
Received 8 May 2023; Received in revised form 13 October 2023; Accepted 9 November 2023
Available online 18 November 2023
2352-152X/© 2023 Elsevier Ltd. All rights reserved.
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
enhanced. Meanwhile, a hybrid tram simulation model is established to (2) Optimization-based methods
verify the effectiveness of energy management strategies. The contri
butions of our work are as follows. Firstly, the energy management Optimization-based methods include global optimization and tran
strategy optimization models based on single threshold and multiple sient optimization [9].
thresholds are constructed by taking energy consumption of power The global optimization method is for determining the operating
battery as objective function and regarding the end voltage, discharge conditions of the vehicle, under multiple constraints, with cost, energy
current and charge state as constraints. Secondly, an improved particle consumption, and efficiency as the optimization objectives, with charge
swarm optimization (PSO) algorithm with competitive mechanism and state of the energy storage element, and output power as the optimiza
dynamic inertia weights is developed to obtain the optimal energy tion variables, using optimization algorithms for global optimization,
management strategy. Thirdly, Guangzhou Haizhu tram is used to verify and the global optimization method is usually combined with the rule-
the performance of the developed method, and the improved PSO based energy management strategy. Yuan et al. [18] used the
method is compared with other two methods to illustrate the reliability. maximum output power of each energy storage element and super
The rest of this paper is organized as follows. Some references on capacitor charge state as the threshold value, and the minimum
energy management strategy optimization are introduced in Section 2. configuration cost of energy storage elements and operational charac
Then, the energy management strategy optimization models and solu teristics index as the multi-optimization objectives, and used a multi-
tion methods based on improved PSO algorithms are developed in objective genetic algorithm (GA) to optimize the threshold value to
Section 3. In Section 4, Guangzhou Haizhu tram is used to verify the achieve the optimal configuration scheme and energy management
performance of the developed method, and the results are presented. strategy of the energy storage system. Wang et al. [19] used the energy
Finally, some conclusions are made in Section 5. management strategy of dynamic power ratio to solve the optimal power
ratio using the PSO algorithm with the cost, quality, and lifetime of
2. Literature review energy storage components as the comprehensive optimization objec
tives. Li et al. [20], Zhang et al. [21], and Benmouna et al. [22] regarded
The energy management strategy of HESS is to reasonably allocate the HESS loss minimization as the optimization objective, discrete the
the work tasks of different energy storage components, mainly including battery charge state, and used the dynamic programming algorithm
rule-based energy management strategy, optimization-based energy under the constraints to obtain the optimal energy management strat
management strategy, and power prediction-based energy management egy. The global optimization method is developed under a defined route
strategy. and specific operating conditions and is not suitable for the case of
variable operating conditions; however, it has a good optimization effect
(1) Rule-based methods when used to optimize other energy management strategies.
The transient optimization method is a method to solve the optimi
The rule-based methods include deterministic rules and fuzzy rules zation variables using optimization algorithms for real-time optimiza
[9]. tion objectives. Few studies on transient optimization have been
Deterministic rules are mainly based on logical threshold methods, i. conducted, Hu et al. [23] and Anbarasu et al. [24] used the minimum
e., setting a certain parameter as a threshold and comparing it with the equivalent fuel consumption as the optimization objective to change the
real-time state of the train to determine the operating state of each en battery output power profile by adjusting the fuel cell output power in
ergy storage element. Yedavalli et al. [10] and Guo et al. [11] used real-time to achieve optimal power allocation. The real-time optimiza
supercapacitor charge state as a fixed threshold, combined with the tion algorithm is less computationally intensive compared to the global
vehicle demand power, to determine the power supply and recovery optimization algorithm and can achieve the real-time optimum, but
state of the energy storage system, and then determined the working cannot guarantee the global optimum.
state of the energy storage elements. Guo et al. [12] and Graf et al. [13]
used the maximum output power of the power battery and the maximum (3) Power prediction-based methods
input power of the supercapacitor as fixed thresholds and combined
with the vehicle demand power to determine the working state of energy The energy management strategy based on power prediction is to
storage elements. The above fixed threshold methods are developed predict the future power considering the current and future working
based on the determination of the operating state of the train, and the conditions to assign the working state of energy storage elements. Zhang
threshold values are determined by experts' experience. Fixed threshold et al. [25] proposed an energy management strategy based on acceler
methods have been improved, Wang et al. [14] used the fixed power ation time prediction, the maximum power available for the super
ratio as the threshold value, optimized the original fixed power ratio, capacitor was calculated during the vehicle acceleration time, and the
and reduced the number of energy storage elements by considering the actual maximum discharge power was used as the power threshold to
energy interaction between supercapacitor and power battery under the allocate power battery and supercapacitor power. Peng et al. [26] pro
target function of energy storage element quality. Shen et al. [15] pro posed an energy management strategy that considers the power demand
posed a dynamic threshold energy management strategy. For different at the before and after moments. On the determined line, the power
demand powers of the power battery in different intervals, the power zones are divided, the energy consumption of the before and after power
battery discharge current with the goal of minimizing the power battery zones are estimated, and the maximum performance satisfaction rate
energy consumption was optimized to dynamically adjust the power and the maximum energy absorption rate are used as evaluation indexes
battery output power threshold value in each interval. The deterministic to find the optimal target power curve. Shen et al. [27] and Xie et al.
rules method has the advantages of deterministic rules, practicality, and [28] developed Markov Monte Carlo models to predict the power de
easy implementation. It also has the disadvantage of poor adaptability mand and optimize the power allocation under the current moment and
and inability to achieve optimal system energy distribution. the future moment to achieve the optimal objective function under the
Fuzzy rules mainly use fuzzy control methods to fuzzify control rules current and future moments. The energy management strategy based on
and select appropriate input variables. Gao and Gao [16] used the power prediction can adapt to the changes in working conditions, but
available energy of power battery, supercapacitor, and vehicle demand the calculation is complex and difficult to solve, and there are still some
power as input variables. Trinh et al. [17] used the charge state of the problems in vehicle applications.
energy storage element and vehicle demand power as input variables. In summary, the energy management strategies applied to trams are
Fuzzy rules have good adaptability, but fuzzy rules can cause large ac mainly developed for specific lines and defined vehicles, while single
curacy errors. energy management strategies still have some limitations. Therefore,
2
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
where t0 and t1 are the start and end time of power battery energy supply
in the k-th interval, Ubk , Ibk and Ebk are the end voltage, discharge current
and energy consumption of the power battery at the k-th interval, and
k = 1, 2, ⋯10.
In the iteration process, the power battery and supercapacitor should
supply energy in normal working conditions, and the constraints are
⎧
⎪
⎪ Ub.min ≤ Ub ≤ Ub.max
⎪
⎪
⎪
⎪ Uc.min ≤ Uc ≤ Uc.max
⎨
Ib.min ≤ Ib ≤ Ib.max
(2)
⎪
⎪ Ic.min ≤ Ic ≤ Ic.max
⎪
⎪
⎪
⎪ SOCmin ≤ SOC ≤ SOCmax
⎩
SOEmin ≤ SOE ≤ SOEmax
Fig. 2. Power allocation based on multiple thresholds.
where Ub and Uc are the end voltage of power battery and super
the combination of different energy management strategies will be the capacitor, Ib and Ic are their discharge current, and SOC and SOE are
focus of future research, and specific combination schemes can be their charge state.
determined according to the characteristics of the research objects and
research objectives. 3.1.2. Multiple thresholds
The different line conditions cause the variation of DC bus demand
3. Methodology power. If the same power threshold is used for the whole line, it will
inevitably lead to differences in the power supply capacity of the
Efficient energy management is one of the key points for the energy supercapacitor at each interval.
storage system to utilize its power supply capacity and meet the oper To solve the above problem, the power threshold of each interval is
ational needs of the tram. Based on the optimal parameter configuration optimized separately by combining the difference of DC bus demand
of the HESS, the energy management strategy with a fixed power power among stations. Multiple thresholds Pbto exist in a whole oper
threshold value is optimized from the perspective of the whole line and ating line, and the power allocation is shown in Fig. 2. Each operation
multiple intervals. The energy management strategy optimization in interval has an optimal power threshold, Pbto1 , Pbto2 and Pbto3 represent
cludes single threshold values and multiple threshold values. the optimal power threshold values of the first, second, and third in
terval, respectively.
3.1. Energy management strategy optimization For the optimization of multiple power thresholds, each interval is
regarded as an optimization interval, the minimum energy consumed by
To maintain the strong range of the power battery, the energy supply the power battery in each interval is the optimization objective, the
distribution of the energy storage components needs to be adjusted so power battery discharge current in each interval is the optimization
that the supercapacitor can assume more power supply to reduce the control variable, and an improved PSO algorithm is used to obtain the
energy consumption of the power battery and improve its state of optimal threshold values for each interval in the energy supply stage of
charge. the energy storage system.
The energy management strategy of fixed power thresholds requires In summary, the optimization objective function F2 satisfies
the adjustment of power thresholds. The power supply phase requires ( ∫ t1 )
lowering the power threshold, i.e., lowering the output maximum power minF2 (Ibk ) = min(Ebk (Ibk ) ) = min Ubk Ibk dt (3)
Pbmo of the power battery. The recovery energy phase requires main
t0
taining the supercapacitor charging capacity, i.e., maintaining the The energy storage element should be worked in the normal oper
maximum supercapacitor charging power Pcmi . The output maximum ating state in the multi-power threshold optimization process. The
power Pbmo of the power battery is obtained by multiplying the instan voltage, current, and charge state should satisfy the constraints of Eq.
taneous terminal voltage Ub and the maximum output current Ibmo , and (2).
the variation of terminal voltage Ub is affected by Ibmo , thus Pbmo can be
adjusted by adjusting Ibmo . 3.2. Solution method based on improved PSO algorithm
3
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
and Ib.min are the maximum and minimum values of power battery
discharge current, respectively.
4
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
Table 1 smaller inertia weight w is used to perform a better local search; for
Parameter settings. particles far from the population optimal solution, a larger inertia
Parameters Values weight w is used to perform a better global search. Considering the
distance of the particles from the population optimal solution, the
Vmax 2
Vmin − 2 dynamically changing inertia weight wd satisfies [35]:
wmax 0.8 ( t )
wmin 0.3
Li − Lti.min (wmax − wmin )
wd = wmin − ( t ) (6)
c1 , c 2 1 Li.max − Lti.min
Maximum iteration number n 80
Consecutive iteration number of optimal fitness value m 5 where Lti is the distance between the i-st particle and Gbest t in the t-th
Population number N 7
Current fixed step ΔIb 5
iteration; Lti.min and Lti.max are the minimum and maximum values of Lti ,
α, β 0.4, 0.6 respectively; i is the particle number and satisfies i = 1, 2, ⋯N.
The inertia weight wn decreases with the iteration number t in
creases, and the integrated inertia weight w is composed of wd and wn ,
Table 2 and satisfies
Basic parameters of energy storage elements. {
w = αwd + βwn
(7)
Parameters Values α+β=1
SOCmax 90 %
SOCmin 30 % where α and β are constants.
SOEmax 100 % The following steps of the improved PSO algorithm are presented.
SOEmin 25 %
Ibmin − 68A
Step 1: Input the effective particles and determine the number N of
Ibmax 35A
Icmin − 450A effective particles.
Ibmax 450A Step 2: Determine the fitness function. F1 and F2 are the fitness
functions of single threshold and multiple thresholds, respectively.
Step 3: Calculate the particle fitness value. Under the corresponding
population position average is considered to maintain the global search
power threshold value of each particle, the energy consumption of
capability when updating the velocity of the failed particles.
the power battery is calculated during the power supply simulation,
For a population consisting of N particles, two particles are randomly
i.e., the fitness value.
assigned to form a pair, and the assigned particles are withdrawn from
Step 4: Randomly assign particle pairs and rule the winning and
the population to be paired. If N is an even number, N/2 pairs are
losing particles. Two particles are randomly assigned to form a
formed; if N is an odd number, a particle is randomly taken from the N −
particle pair and the fitness values of two particles are compared. The
1 previous paired particles, and (N + 1)/2 pairs are formed [34]. The
particle with a smaller fitness value is the winning particle and the
energy consumption of the power battery is used as the fitness function.
particle with a larger fitness value is the losing particle.
The particle with a smaller fitness value is used as the winning particle
Step 5: Update the particle velocity and position. The inertia weight
and the particle with a larger fitness value is used as the losing particle.
w is changed by Eq. (7), and then the velocity and position of the
For each iteration process, the winning and losing particles satisfy
winning and losing particles are updated by Eq. (5).
the following requirements.
Step 6: Exit the optimization solving algorithm and output the
⎧
⎪ particles.
⎪
⎪ t+1 t
⎪
⎪ VW,k = wVW,k
⎪
⎪
⎪
⎪
⎪ t+1
XW,k = wXW,kt 4. Results
⎪
⎪
⎪
⎪ ( ) ( )
⎨ V t+1 = wV t + c r X t − X t t t
L,k L,k 1 1 W,k L,k + c2 r2 X − XL,k
(5) Guangzhou Haizhu tram is used to verify the performance of the
⎪
⎪
⎪
⎪
t+1
XL,k t
= XL,k t+1
+ VL,k developed method [30]. Assume that the HESS has an optimal config
⎪
⎪
⎪ uration of energy storage elements, Nbs = 221, Nbp = 10, Ncs = 125, and
⎪
⎪ Vmin ≤ Vit+1 ≤ Vmax
⎪
⎪
⎪
⎪ Ncp = 10. The whole line and each interval are taken as the optimization
⎪ Xmin ≤ Xit+1 ≤ Xmax
⎩ intervals, and the optimal discharge current is finally obtained. The
parameter settings of the developed algorithm are shown in Table 1 and
t
where VW,k t
and XW,k are velocity and position of the winning particle in the parameters of energy storage elements are shown in Table 2.
t
the t-th iteration of the k-th pair; VL,k t
and XL,k are the velocity and po Taking the energy management strategy optimization based on a
single threshold value as an example, the iterative solutions based on the
sition of the losing particle; Xt is the average position of the population classical PSO algorithm and the improved PSO algorithm are used
in the t-th iteration; k is the particle pair number, when N is an even respectively, and their corresponding iterative changes of the optimal
number, then k = 1, 2, ⋯N/2, when N is an odd number, then k = 1, 2, discharge current are shown in Fig. 4. The optimal discharge current
⋯(N + 1)/2. converges to 23.98A under both two algorithms. Compared with the
classical PSO algorithm, the improved PSO algorithm has a better search
(2) Dynamic inertia weights capability at the beginning of the iteration. At the end of the iteration, it
has a faster convergence rate.
If the particle swarm searches for the optimal solution at the For the energy management strategy optimization based on single
beginning of the iteration, a larger inertia weight w will cause the swarm threshold and multiple thresholds, the improved PSO algorithm is used
to jump around the optimal solution at a larger speed, which slows down to obtain the optimal discharge current of the power battery from the
the convergence speed; if the swarm diversity decreases at the end of the perspective of the whole line and each interval, as shown in Fig. 5. For
iteration, a smaller inertia weight w will cause the swarm to fall into a the whole line, the power battery discharge current is fixed at 35A
local optimal solution. The distance between particles and the popula before optimization; under the energy management strategy with the
tion optimal solution Gbestt is calculated and used as the basis for single threshold, the optimal power battery discharge current is 23.98A,
determining w. For particles close to the population optimal solution, a
5
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
6
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
strategies can meet the DC bus demand power, and the power threshold minimum charge states of the supercapacitor under the two strategies
value is significantly reduced compared with that before optimization. are equal and both reach the allowable discharge limit (25 %). In the
The charge state curves for the power supply of the energy storage other intervals, the minimum charge state of the supercapacitor under
elements are shown in Fig. 7. The power battery charge state decreases the multiple thresholds is lower, which indicates that the supercapacitor
continuously with the increase of supply distance, while the super discharges more deeply and has more power supply capability.
capacitor charge state in each interval first supply decreases, then re The power consumption curves of the energy storage elements are
covers the braking energy back up, and charges to full charge at the shown in Fig. 8. From the perspective of the whole operation line, the
station. For the whole operation line, the power battery charge state power consumption curve of the power battery monoblock is flatter
changes as follows: under the single threshold, it decreases from 90 % to under the multiple thresholds, and the power consumption is smaller at
50.42 %, which is higher than the lowest power battery charge state the end of the power supply. The power consumption curves of the
(33.12 %) before optimization, and the optimized power battery charge supercapacitor monoblock are similar under the two optimization stra
state is obviously improved; under the multiple thresholds, it decreases tegies, and the power consumption of the two strategies is identical in
from 90 % to 73.78 %. In the fourth, fifth, eighth, and tenth intervals, the the first interval. In the other intervals, the maximum power consump
charge state remains unchanged and the power battery charge state is tion of the supercapacitor under multiple thresholds is significantly
further improved compared to the single threshold. larger. It indicates that the supercapacitor takes on more power supply
The minimum supercapacitor charge state in each interval under the tasks and plays a larger power supply capacity under multiple
two optimization strategies is shown in Table 3. In the first interval, the thresholds.
7
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
Table 3 Table 4
Minimum charge state of supercapacitor under two optimization strategies. Comparison of electrical parameters.
Intervals Single threshold Multiple thresholds Electrical parameters Single Multiple
threshold thresholds
1 25.00 % 25.00 %
2 48.26 % 34.38 % Minimum charge state of power battery 50.42 % 73.78 %
3 48.81 % 31.68 % Minimum charge state of supercapacitor 25 % 25 %
4 50.47 % 33.51 % Average minimum charge state of 42.97 % 31.87 %
5 50.64 % 33.48 % supercapacitor
6 41.16 % 25.11 % Total energy consumption of power battery 17.54 kWh 7.35 kWh
7 29.20 % 25.05 % pack
8 56.98 % 45.43 % Total energy consumption of supercapacitor 33.34 kWh 43.53 kWh
9 25.62 % 25.01 % bank
10 53.55 % 40.00 %
8
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
(a) Power battery SOC under single threshold (b) Power battery SOC under multiple thresholds
(c) Minimum SOE of supercapacitor under single threshold (d) Minimum SOE of supercapacitor under multiple thresholds
(e) Energy consumption of power battery under single threshold (f) Energy consumption of power battery under multiple thresholds
Fig. 10. The end voltage of the power battery monoblock decreases as supercapacitor discharge capacity is greater under the multiple
the supply distance increases, and comparing the two optimization thresholds.
strategies, the end voltage changes more slowly and the end voltage The electrical parameters under the two optimization strategies are
reduction is smaller for the multiple thresholds. The end voltages of compared in Table 4. The average minimum charge state of the super
supercapacitors have similar trends under the two optimization strate capacitor is the average value of the minimum charge state of the
gies. Except for the first, seventh, and ninth intervals which are basically supercapacitor in 10 intervals. Comparing the two optimization strate
the same, the minimum value of end voltage is smaller under the mul gies, at the end of the power supply, under the multiple thresholds, the
tiple thresholds in all other intervals. Combining the voltages and cur minimum charge state of the power battery is improved by 23.36 %, and
rents, it shows that the energy storage element works in normal the total energy consumption of the power battery pack is reduced by
conditions under both two optimization strategies, and the 58.10 %, and the average minimum charge state of supercapacitor is
9
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
reduced by 11.10 %, and the total energy consumption of supercapacitor CRediT authorship contribution statement
pack is improved by 30.56 %. Therefore, the energy management
strategy optimization based on multiple thresholds has a more obvious Zhenyu Zhang: Methodology, Software, Visualization, Writing –
effect in reducing power battery energy consumption and exploiting the original draft. Xiaoqing Cheng: Data curation, Writing – original draft.
power supply capacity of supercapacitors. Zongyi Xing: Supervision, Writing – review & editing. Zihao Wang:
In addition, the optimal energy management strategy obtained by Conceptualization, Data curation, Writing – original draft.
improved PSO methods is compared with those of GA [18] and PSO [19]
methods, and the comparative results of the three methods are shown in Declaration of competing interest
Fig. 11. Then, some conclusions are provided as follows.
Firstly, the improved PSO method improves the minimum charge The authors declare that they have no known competing financial
state of power battery. Under the single threshold, the minimum charge interests or personal relationships that could have appeared to influence
states of power battery obtained by GA, PSO, and improved PSO the work reported in this paper.
methods are 41.13 %, 41.62 %, and 50.42 % (Fig. 11(a)), all higher than
before optimization. Under multiple thresholds, the minimum charge Data availability
states of the three methods are 72.00 %, 72.31 %, and 73.78 % (Fig. 11
(b)), respectively. Both single threshold and multiple thresholds, the No data was used for the research described in the article.
improved PSO method has better performance of power battery than GA
and PSO methods. Acknowledgement
Secondly, the improved PSO method can more effectively utilize the
power supply capacity of supercapacitors. Under the single threshold, This work was supported by the National Natural Science Foundation
the average supercapacitor minimum charge state of all the 10 intervals of China (No. 52302436) and the State Key Laboratory of Rail Traffic
obtained by the three methods are 47.56 %, 47.60 %, and 42.97 % Control and Safety (Contract No. RCS2023K004), Beijing Jiaotong
(Fig. 11(c)). Under multiple thresholds, the average supercapacitor University.
minimum charge states by the three methods are 35.27 %, 35.31 %, and
31.87 % (Fig. 11(d)). Both single threshold and multiple thresholds, the References
improved PSO method enables supercapacitors to provide more energy
for tram operation. [1] M. Steczek, W. Jefimowski, T. Maciołek, S. Barański, Analysis of energy losses in
the novel distributed power supply system for trams, IEEE Trans. Transp.
Thirdly, the improved PSO method decreases the energy consump Electrification 9 (1) (2022) 1498–1508.
tion of the power battery. Under the single threshold, the total energy [2] Z. Xing, Z. Zhang, J. Guo, Y. Qin, L. Jia, Rail train operation energy-saving
consumption of power batteries by the three methods is 35,303 J, 34946 optimization based on improved brute-force search, Appl. Energy 330 (2023),
120345.
J, and 28,590 J (Fig. 11(e)). Under the single threshold, the total energy [3] J.J. Mwambeleko, T. Kulworawanichpong, Supercapacitor and accelerating
consumption by the three methods is 13,234 J, 13240 J, and 11,980 J contact lines hybrid tram system, J. Energy Storage 44 (2021), 103277.
(Fig. 11(f)). Both the two thresholds, the improved PSO method makes [4] J.J. Mwambeleko, T. Hayasaka, T. Kulworawanichpong, Enhancing conventional
battery and contact line hybrid tram system with accelerating contact lines, IET
power battery supply less energy for tram operation. Electr. Syst. Transp. 10 (1) (2020) 105–115.
[5] Z. Zhang, H. Zhang, L. Zhou, Zero-carbon measure prioritization for sustainable
5. Conclusions freight transport using interval 2 tuple linguistic decision approaches, Appl. Soft
Comput. 132 (2023), 109864.
[6] Y. Wang, L. Wang, M. Li, Z. Chen, A review of key issues for control and
This paper presents an energy management strategy optimization management in battery and ultra-capacitor hybrid energy storage systems,
method of HESS based on a fixed power threshold. The innovations of ETransportation 4 (2020), 100064.
this paper are provided as follows. [7] Z. Zhang, H. Zhang, L. Zhou, Y. Qin, L. Jia, Incomplete Pythagorean fuzzy
preference relation for subway station safety management during COVID-19
Firstly, to address the problem of energy management strategy pandemic, Expert Syst. Appl. 119445 (2022).
optimization with a fixed power threshold, the threshold values are [8] W. Yang, J. Ruan, J. Yang, N. Zhang, Investigation of integrated uninterrupted dual
optimized from the perspective of the whole line and each interval, and input transmission and hybrid energy storage system for electric vehicles, Appl.
Energy 262 (2020), 114446.
the energy management strategy optimization models based on the [9] S.F. Tie, C.W. Tan, A review of energy sources and energy management system in
single threshold and multiple thresholds are constructed by taking en electric vehicles, Renew. Sust. Energ. Rev. 20 (2013) 82–102.
ergy consumption of power battery as objective function and regarding [10] K. Yedavalli, L. Guo, D.S. Zinger, Simple control system for a switcher locomotive
hybrid fuel cell power system, IEEE Trans. Ind. Appl. 47 (6) (2011) 2384–2390.
the end voltage, discharge current and charge state as constraints. [11] L. Guo, K. Yedavalli, D. Zinger, Design and modeling of power system for a fuel cell
Secondly, the competitive PSO algorithm with dynamic inertia hybrid switcher locomotive, Energy Convers. Manag. 52 (2) (2011) 1406–1413.
weights has a faster convergence rate and can maintain the diversity of [12] Y. Guo, X. Dai, K. Jermsittiparsert, N. Razmjooy, An optimal configuration for a
battery and PEM fuel cell-based hybrid energy system using developed Krill herd
the swarm population, which is developed to solve optimization models, optimization algorithm for locomotive application, Energy Rep. 6 (2020) 885–894.
and the energy management strategy optimization algorithm is [13] T. Graf, R. Fonk, J. Schröter, P. Hoenicke, C. Bauer, J. Kallo, C. Willich,
designed. Investigation of a fuel cell hybrid system with a new modular test bench approach
for all electric hybrid power train systems, J. Energy Storage 56 (2022), 105999.
Thirdly, Guangzhou Haizhu tram is used to illustrate the perfor
[14] X. Wang, Y. Luo, B. Qin, L. Guo, Power allocation strategy for urban rail HESS
mance of the developed method. The results show that the optimal en based on deep reinforcement learning sequential decision optimization, IEEE
ergy management strategy based on multiple thresholds has better Trans. Transp. Electrification 9 (2) (2023) 2693–2710.
effects than single threshold in giving full play to the power supply ca [15] Y. Shen, P. Cui, X. Wang, X. Han, Y.X. Wang, Variable structure battery-based fuel
cell hybrid power system and its incremental fuzzy logic energy management
pacity of supercapacitor and reducing the energy consumption of power strategy, Int. J. Hydrog. Energy 45 (21) (2020) 12130–12142.
battery. The minimum charge state of the power battery under multiple [16] F. Gao, X. Gao, Hyper-spherical search optimized fuzzy logic control considering
thresholds is improved by 23.36 % over that of single threshold, and the operating conditions for hybrid tram, IEEE Access 10 (2022) 65925–65935.
[17] H.A. Trinh, H.V.A. Truong, K.K. Ahn, Development of fuzzy-adaptive control based
total energy consumption of the power battery pack is reduced by 58.10 energy management strategy for PEM fuel cell hybrid tramway system, Appl. Sci.
%. Meanwhile, comparing with GA and PSO methods, the improved PSO 12 (8) (2022) 3880.
method effectively improves the minimum charge state of power bat [18] Y. Yuan, M. Chen, J. Wang, W. Yu, B. Shen, A novel hybrid energy management
strategy of a diesel-electric hybrid ship based on dynamic programing and model
tery, utilizes more energy from supercapacitors, and decreases the en predictive control, Proc. Inst. Mech. Eng. Part M: J. Eng. Marit. Environ.
ergy consumption of the power battery. 14750902211068931 (2022).
10
Z. Zhang et al. Journal of Energy Storage 75 (2024) 109698
[19] Y. Wang, Z. Yang, F. Lin, X. An, H. Zhou, X. Fang, A hybrid energy management [27] D. Shen, C.C. Lim, P. Shi, P. Bujlo, Energy management of fuel cell hybrid vehicle
strategy based on line prediction and condition analysis for the hybrid energy based on partially observable Markov decision process, IEEE Trans. Control Syst.
storage system of tram, IEEE Trans. Ind. Appl. 56 (2) (2020) 1793–1803. Technol. 28 (2) (2018) 318–330.
[20] Q. Li, B. Su, Y. Pu, Y. Han, T. Wang, L. Yin, W. Chen, A state machine control based [28] S. Xie, H. He, J. Peng, An energy management strategy based on stochastic model
on equivalent consumption minimization for fuel cell/supercapacitor hybrid predictive control for plug-in hybrid electric buses, Appl. Energy 196 (2017)
tramway, IEEE Trans. Transp. Electrification 5 (2) (2019) 552–564. 279–288.
[21] G. Zhang, Q. Li, W. Chen, X. Meng, H. Deng, A coupled power-voltage equilibrium [29] X. Cai, H. Qiu, L. Gao, C. Jiang, X. Shao, An efficient surrogate-assisted particle
strategy based on droop control for fuel cell/battery/supercapacitor hybrid swarm optimization algorithm for high-dimensional expensive problems, Knowl.-
tramway, Int. J. Hydrog. Energy 44 (35) (2019) 19370–19383. Based Syst. 184 (2019), 104901.
[22] A. Benmouna, M. Becherif, L. Boulon, C. Dépature, H.S. Ramadan, Efficient [30] Z. Xing, J. Zhu, Z. Zhang, Y. Qin, L. Jia, Energy consumption optimization of
experimental energy management operating for FC/battery/SC vehicles via hybrid tramway operation based on improved PSO algorithm, Energy 258 (2022),
artificial neural networks-passivity based control, Renew. Energy 178 (2021) 124848.
1291–1302. [31] Z. Zhang, X. Cheng, Z. Xing, Z. Wang, Y. Qin, Optimal sizing of battery-
[23] X. Hu, C. Zou, X. Tang, T. Liu, L. Hu, Cost-optimal energy management of hybrid supercapacitor energy storage systems for trams using improved PSO algorithm,
electric vehicles using fuel cell/battery health-aware predictive control, IEEE J. Energy Storage 73 (2023), 108962.
Trans. Power Electron. 35 (1) (2019) 382–392. [32] K. Liu, X. Wang, Z. Qu, Train operation strategy optimization based on a double-
[24] A. Anbarasu, T.Q. Dinh, S. Sengupta, Novel enhancement of energy management in population genetic particle swarm optimization algorithm, Energies 12 (13) (2019)
fuel cell hybrid electric vehicle by an advanced dynamic model predictive control, 2518.
Energy Convers. Manag. 267 (2022), 115883. [33] C. Huang, X. Zhou, X. Ran, Y. Liu, W. Deng, W. Deng, Co-evolutionary competitive
[25] Z. Zhang, H. He, J. Guo, R. Han, Velocity prediction and profile optimization based swarm optimizer with three-phase for large-scale complex optimization problem,
real-time energy management strategy for plug-in hybrid electric buses, Appl. Inf. Sci. 619 (2023) 2–18.
Energy 280 (2020), 116001. [34] X. Xue, P.W. Tsai, Integrating energy smart grid’s ontologies through multi-
[26] H. Peng, H. Cao, S. Dirkes, Z. Chen, K. Deng, J. Gottschalk, K. Hameyer, Validation objective particle swarm optimization algorithm with competitive mechanism,
of robustness and fuel efficiency of a universal model-based energy management Sustain. Energy Technol. Assess. 53 (2022), 102442.
strategy for fuel cell hybrid trains: from analytical derivation via simulation to [35] S. Wang, G. Liu, M. Gao, S. Cao, A. Guo, J. Wang, Heterogeneous comprehensive
measurement on test bench, Energy Convers. Manag. 229 (2021), 113734. learning and dynamic multi-swarm particle swarm optimizer with two mutation
operators, Inf. Sci. 540 (2020) 175–201.
11