Keywords

1 Introduction

In recent years, the gas emitted by traditional fuel vehicles has aggravated air pollution and fossil energy is non-renewable. Therefore, it’s essential to develop and enter a period of accelerated development of electric vehicles (EV).

In [1], it indicates that high discharge rates and depth of discharge will cause high rate of capacity fade and a shorter useful life of lithium-ion battery (LiB). Studies have shown that when the LiB faces a high-power impact, it may lead to over-discharge of the LiB, leading to the reduction of the number of charge and discharge cycles of the LiB. Ultracapacitors (UC) can meet the high-power requirements of EV, but their low energy density makes it difficult for vehicles to have a long life. The hybrid energy storage system (HESS) composed of LiB and UC plays a role of "peak cutting and valley filling" for LiB. In [2], the results show that HESS with appropriate size and enabled energy management can significantly reduce the battery degradation rate by about 40% compared to battery energy storage systems (ESS), and at only 1/8 of the additional cost of the system.

HESS primarily encompass three types of topologies: passive topology, semi-active topology, and active topology [3]. Pratim Bhattacharyya et al. [4] proposed an improved LiB and UC hybrid semi-active structure for EVs where the size and space of the energy storage system (ESS) are critical. Zhu et al. [5] proposed that bidirectional DC/DC converters can be divided into isolated (IBDC) and non-isolated bidirectional DC/DC converters (NBDC) according to whether electrical isolation is realized between ports. Li et al. [6] conducted an optimization design for a UC-based semi-active HESS and interleaved parallel bidirectional Buck/Boost converters, and analyzed the reasons behind the optimization results.

A good energy management strategy (EMS) of HESSs can improve performance in different and complex driving conditions and reduce driving costs. HESS energy management methods [7,8,9] in EV applications can be summarized into the following two categories, namely model-based methods and rule-based methods. Model-based energy management methods usually cooperate with other control methods for power distribution, such as the global optimization algorithm dynamic programming (DP) [10], which can obtain the optimal control input by minimizing the cost function, and it’s suitable for offline calculation to provide parameter setting values for determining rule control. After analyzing different data sets according to the optimal power distribution strategy, Shen et al. [11] proposed a neural network-based (NN-based) HESS control method for medium-sized EVs. Rule-based methods consist of a set of predefined, empirical control rules [12]. Wang et al. [13] proposed a rule-based control method to realize mode selection and current distribution of multi-mode HESS in EV applications. Yin et al. [14] proposed an adaptive fuzzy logic control (FLC) scheme for EV energy management, where the output membership functions are periodically refreshed to adapt to changing driving patterns. The experimental results in An et al. [15] conducted under the HWFET condition demonstrate that the fully active dual-energy source HESS, along with the EMS based on FLC, effectively safeguards the LiB against the detrimental effects of substantial current fluctuations, consequently prolonging the battery’s lifespan. Based on the advantages of flexibility and robustness of the fuzzy logic controller, Gao et al. [16] proposed an optimization method of fuzzy controller hybridization (DOH) and membership function based on the golden section rule. In [17], The optimal logical threshold control (LTC) can fully leverage the high-power UC’s characteristics and conduct hardware-in-the-loop (HIL) experiments to further validate the real-time and dependable nature of the near-optimal LTC. The gray wolf optimization (GWO) is used in Hu et al. [18] to optimize the battery output power upper limit and UC charging upper limit of mature multi-mode control (MMC). In [19], frequency-based energy management distributes high- and low-frequency power requirements to batteries and ultracapacitors respectively. Min [20] proposed a multi-objective EMS for EV HESS based on separating load power. The impact of different sorting methods on the results by using elite strategy non-dominating sorting genetic algorithms (NSGA-II).

With the widespread adoption of EVs, experts and scholars have initiated research into artificial intelligence power distribution strategies. On the basis of existing algorithms, it’s necessary to improve the robustness of EV under driving conditions and optimize the control strategy algorithm. The sections of this article are as follows. In the second section, we introduce vehicle parameters, HESS components, and the simulation model. Section 3 provides an overview of the LTC and FLC strategies. Sections 4 and 5 present the simulation results and draw conclusions, followed by future prospects.

2 Parameters and Hybrid Energy Storage System Model

2.1 Vehicle Parameters

The parameters of the electric vehicle are designed according to a benchmark model. The main vehicle parameters are shown in Table 1.

Table 1 Main vehicle parameters
Full size table

2.2 Lithium-ion Battery

The equivalent circuit models of LiB include Rint model, RC model, PNGV model, etc. Due to its concise structure and convenient calculation method, the Rint model was selected in this research. There is a certain relationship between the open circuit voltage UOC and the load voltage Ub, as shown in equation: \({U}_{b}={U}_{OC}-{I}_{L}\cdot {R}_{0}\). The parameter settings of the LiBs are shown in Table 2.

Table 2 Parameter settings of LiBs
Full size table

2.3 Ultracapacitor

Under high power requirements, ultracapacitor can provide excellent output power, effectively reduce the load of LiB, and significantly improve the efficiency of the system. The parameter settings of the UCs are shown in Table 3.

Table 3 Parameter settings of UCs
Full size table

2.4 Topology Structure

Figure 1 shows the specific classification of HESS topology. We used a UC semi-active topology in this study for the following reasons: passive topology makes it difficult to achieve energy conversion between LiB and UC, while active topology is more expensive and harder to control. In the semi-active topology of LiB, the linear charge and discharge characteristics of UC can lead to sharp fluctuations in DC circuit voltage. The UC semi-active topology effectively improves capacity efficiency. The formed topology model is presented in Fig. 2.

Fig. 1
A classification chart of H E S S topology segments into active, semi-active, and passive topology. Semi-active topology further segments into L i B and U C semi-active.

Topology structure classification of HESS

Full size image
Fig. 2
A cyclic flow diagram for the U C semi-active topology simulation starts with the storage and ultracapacitor system and flows through the subsystem and D C. In the subsystem, avail from battery option points to storage and avail from ultra capacity points to the ultra-capacitor via power required from control strategy power required from D C.

UC semi-active topology simulation model

Full size image

3 Control Strategy

3.1 Logical Threshold Control Strategy

In this study, writing an integrated program in Matlab to calculate the energy demand according to the required power of the UDDS model of an electric vehicle with a single power supply, to get the Pave_p logic threshold, the related parameters are listed in Table 4. The logic flow of the LTC strategy is presented in Fig. 3.

Table 4 Average power required in UDDS
Full size table
Fig. 3
A flowchart for the logic threshold control strategy starts with P req. If P req is greater than 0, it leads to P a v e p and S O C s c greater than or equal to 0.5, or else, S O C s c less than or equal to 0.997. P bat and P s c equivalents are provided based on the satisfaction of conditions.

Logic threshold control strategy flowchart

Full size image

3.2 Fuzzy Logic Control Strategy

This research employs the Madamni structure fuzzy logic controller, where the input signals consist of requested power, SOCuc, and SOCbat. The output is denoted as the power allocation factor Kbat. The domains and subsets of the membership functions are presented in Table 5.

Table 5 Function domain and subset division
Full size table

By dividing the intervals as outlined above, we derive a total of 108 control rules. The surface of FLC rules under both driving and braking conditions is illustrated in Fig. 4, and its model is further visualized in Fig. 5.

Fig. 4
Two 3-D surface graphs of K bat versus P req and S O C battery in a and K bat versus S O C battery and P req in b. A labeled P req greater than 0, under driving conditions follows a falling trend. B labeled P req less than 0, under braking conditions follows a falling trend that extends as rising.

FLC rule surface under driving and braking conditions

Full size image
Fig. 5
A Simulink model of the F L C starts includes input 1, constant, constant 1, switch and switch 1, S O C and S O C 2 from 1, 2, 3, and 4, fuzzy logic controller and fuzzy logic controller 1, switch 2, product to battery and product to capacity via subtract.

FLC simulation model

Full size image

4 Simulation Results

In this study, we conducted simulation verification within the Matlab and ADVISOR environments to align vehicle parameters and EV control models. Both the LiB and UC were initialized with a State of Charge (SOC) of 1. As depicted in Fig. 6, the speed request curves generated by the two distinct control strategies closely match the actual speed curves. This substantiates that the HESS control strategy proposed in this paper adequately fulfills the vehicle’s speed requirements.

Fig. 6
A line graph of vehicle speed versus time plots fluctuating trends for the required speed under U D D S driving circles, real speed under logical threshold control strategy, and real speed under fuzzy logic control strategy. They pass through (1, 0), (101, 55), (251, 90), (501, 60), (801, 55), and (1351, 35).

Vehicle speed tracking results of two control strategies in UDDS

Full size image

In Fig. 7, within the single-power controlled EMS, the peak output power of the LiB reaches 50.72 kW. This is notably higher, by 40.52 and 39.33 kW, than the values obtained using LTC and FLC strategies in the HESS, as detailed in Table 6. In Fig. 8, the SOC drop of the LiB within the HESS is more than 30% lower compared to the EMS relying on a single power supply. This observation underscores the crucial role of well-distributed power in the HESS, effectively extending the lifespan of the LiB and potentially increasing the mileage range of EVs.

Fig. 7
A line graph of power versus time plots overlapped fluctuating trends for E S S, H E S S, and H E S S based on a single power source, based on logic threshold control strategy, and based on fuzzy logic control strategy, respectively. E S S has the highest at (201, 50000) and the lowest at (101, negative 12000).

Power distribution of LiB based on different energy management strategies

Full size image
Table 6 Output peak and average power of LiB and UC under driving conditions in UDDS
Full size table
Fig. 8
A line graph of S O C of L i B versus time plots falling trends for H E S S based on fuzzy logic control strategy followed by H E S S based on logic threshold control strategy and E S S based on a single power source approximately from (1, 1) to (1301, 0.985), (1301, 0.975), and (1301, 0.965), respectively.

The solutions of LiB SOC in UDDS

Full size image

In LTC, the output power of the LiB is precisely controlled around the threshold value of 10.20 kW, signifying superior output stability of the LiB with a wider amplitude range. In contrast, FLC employs 108 flexible rules to control the LiB’s output power. Since this paper focuses on UC's role in recovering braking energy in LTC, we have compared the average LiB output power under driving conditions in Table 6. In the FLC, the average LiB output power is 2.73 kW lower than that in LTC, highlighting a narrower amplitude range. Figure 7 further illustrates that the LiB’s output power curve exhibits more frequent fluctuations in FLC, indicating inferior stability compared to LTC.

The output power of the UC in the FLC strategy consistently surpasses that of LTC during most driving instances. Additionally, the SOC of the UC in FLC and LTC decreases by approximately 37 and 17%, respectively, signifying a deeper utilization and better alignment of capacity parameters in FLC. However, it’s important to note that this approach may ultimately render the HESS ineffective in the long term, as depicted in Fig. 9.

Fig. 9
A line graph of S O C of U C versus time plots falling trends for H E S S based on logic threshold control strategy followed by H E S S based on fuzzy logic control strategy approximately from (1, 1) to (1301, 0.85) and (1301, 0.65), respectively.

The solutions of UC SOC in UDDS

Full size image

5 Conclusions and Prospect

In this paper, we investigate the EMS for both single power supply and hybrid power supply configurations. The LTC and FLC power distribution strategies effectively regulate the output power of the LiB within an appropriate range. The high-power discharge characteristics of the UC efficiently mitigate excessive LiB discharge, thus ensuring its prolonged service life, exemplifying the significance of the LiB-UC HESS.

In the future, we will consider utilizing the LiB to supply power to the UC in order to meet the power requirements of the UC. Concurrently, further research is required to enhance the alignment of UC capacity parameters and control strategies, incorporating diverse driver behavior patterns and various system topologies into our algorithm design. This will enable us to optimize system costs and control effectiveness while facilitating the flexible selection and application of control strategies based on real-world driving scenarios.