Volume 2017, Issue 13 p. 1356-1361
Article
Open Access

Comparison of EV smart charging strategies from multiple stakeholders' perception

Tanuj Rawat

Corresponding Author

Tanuj Rawat

Department of Electrical Engineering, Malaviya National Institute of Technology, Jaipur, India

Search for more papers by this author
Khaleequr Rehman Niazi

Khaleequr Rehman Niazi

Department of Electrical Engineering, Malaviya National Institute of Technology, Jaipur, India

Search for more papers by this author
First published: 04 December 2017
Citations: 28

Abstract

Electric vehicles (EVs) will become an integral part of the future smart grid. Random charging of EVs may give birth to many issues such as increasing losses, voltage deviation, and increase in peak. The threats imposed by random charging can be conquered by smart or coordinated charging strategies. The liberalisation of energy sector creates an opportunity for different market actors to use flexible EV demand for their own benefits. Thus, the objectives for smart EV charging can be formulated to meet the interest of a single stakeholder or multiple stakeholders. In this study, a comparative analysis of three smart charging strategies from different stakeholders' perception/interests – (i) aggregator (also representing customers), (ii) Network operator, and (iii) both aggregator and network operator simultaneously – while considering different EV penetration is presented in terms of increase in peak load, peak-valley difference, load factor, and total charging cost. The influence of fast charging and battery charging efficiency on these results is also discussed.

1 Introduction

Rising environmental concerns, depleting natural oil, increasing transportation fuel prices, and desire to reduce dependency on imported transportation fuel are some of the reasons why along with renewable energy resources (RES), electric vehicles (EVs) are gaining popularity. In order to reduce greenhouse gas emissions, the governments of many countries are setting targets for EV penetration. For instance, the US government has pledged to achieve 1 million plug-in EVs in the next five years and will contribute US$2 billion incentive for battery development in hybrid EVs [1]. Improvement in battery technology and significant reduction in battery cost together with government policies have compelled different auto-manufacturing companies such as Hyundai, Mahindra, and Renault to roll out EVs to stay in the competitive market [2]. Thus, there will be a large proliferation of EVs in the future smart grid.

Since EVs are propelled by electric motors, they require electrical energy for their operation. The energy demand due to large penetration of EVs will stress the grid. The EV load will also create new load variations in the grid load curve due to the stochastic nature of the driving pattern of users and hence modify the existing grid load profile (without EVs). Several researchers in the past have investigated the effect of uncoordinated (unregulated) charging of EVs. Uncoordinated charging (UCC) causes many issues such as poor load factor, increase in peak-valley difference, and peak load which leads to more losses and voltage deviation [3]. Due to the UCC of EVs, new generation, transmission, and distribution infrastructure would be required as EVs will act as an augmented load for the system, especially during the peak period. According to one survey, most of the EVs are idle 90% of the time. So, EVs can be seen as a flexible load. The EV demand can be controlled such that it aligns with off-peak demand periods and thus can be used to mitigate the challenges of UCC. Therefore, through smart or coordinated charging strategies, EVs load can be used to optimise the grid while satisfying the requirements and constraints of EV owners, resulting in the deferment of grid investments [4]. Smart EV charging strategies will also help to maintain a balance between demand and supply, increase penetration of RES, and improve the efficiency of the system.

Smart charging to make demand profile more uniform by phasing of charging schedule is proposed [5]. In [6], coordinate charging aims to minimise power loss of the distribution system, while different charging periods, 21–6, 18–21, and 10–16 h, are compared. Uncertainty in the arrival time of EV and the initial battery state of charge is not considered [5, 6]. Comparison of different EV charging scenarios while considering the uncertainty of the driving pattern is carried out in [7]. To improve the load profile of system, objectives like minimisation of transformer load profile deviation [8], maximise load factor [9] are also carried out in the literature. Such strategies demonstrate that EV can be integrated into the existing system without increasing the available generating capacity. Comparison of multiple objective functions for smart charging, that is, maximum power objective, minimum cost objective, and maximum wind objective to optimise the charging of EV, is investigated in [10]. Smart charging schemes from both customer and local distribution company's perspective were compared with the uncontrolled charging scheme by Sharma et al. [11].

It is shown in [12] that for a single feeder distribution system, minimising losses maximises the load factor and minimising load variance minimises losses in the context of coordinated EV charging. Maximum EV penetration that can be integrated into a distribution system without undergoing infrastructure reinforcement will also depend on the objective of smart charging considered [13]. While an extensive range of possible smart charging strategies has been proposed in the literature which is applicable for both centralised and de-regulated power system, the optimisation is usually carried out from a single perspective with a single objective. The objectives considered were aimed at maximising profit to either EV owners or network operator. The misalignment of these two objectives shows that coordinated charging problem is complex as it involves multiple stakeholders and each has their own individual interests. Therefore, in this paper, a comparative analysis of three smart charging strategies from different stakeholders' perception – (i) aggregator (also representing customers), (ii) network operator, and (iii) both aggregator and network operator simultaneously – while considering different EV penetration is presented. The comparison is carried out in terms of increase in peak load, peak-valley difference, load factor, and total charging cost. The influence of fast charging and battery charging efficiency on these results is also discussed.

The remaining sections are organised as follows – Section 2 presents the EV driving behaviour model. Mathematical formulation of smart strategies is detailed in Section 3. Section 4 shows results, and finally, Section 5 draws conclusion from this paper.

2 EV model based on driving behaviour

Load modelling of EV is highly stochastic as it needs the driving pattern of EV users. Arrival time, departure time, distance travelled by an EV user, charging rate etc. are some of the parameters required to model the EV demand. Historical data of the driving pattern of users can be studied and probability distribution functions (PDFs) of parameters required for EV modelling can be generated. Based on statistical data of the driving pattern obtained from the National Household Travel survey (NHTS), the daily distance travelled by an EV user will follow the lognormal distribution. Its PDF can be expressed as [9]
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0002(1)
where s is the daily distance travelled by an EV user in miles, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0004 is the mean, and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0006 is the standard deviation of the lognormal distribution. In this case, μ s and σ s are 2.7 and 0.6, respectively.
The distribution of arrival time of an EV will follow the normal distribution. Its PDF can be expressed as [9]
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0008(2)
where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0010 is the arrival time, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0012 is the mean, and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0014 is the standard deviation of the normal distribution. In this case, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0016 and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0018 are18 and 1, respectively. Similarly, the distribution of departure time of an EV from home will also follow the normal distribution. Its PDF can be expressed as [9]
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0020(3)
where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0022 is the departure time, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0024 is the mean, and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0026 is the standard deviation of the normal distribution. In this case, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0028 and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0030 are 7 and 1, respectively.

3 Problem formulations

Objective function for smart charging of an EV can be developed considering the interest of a single stakeholder or multiple stakeholders. In this paper, smart charging strategies considering the interest of different stakeholders, (i) aggregator (also representing customers), (ii) network operator, and (iii) both aggregator and network operator simultaneously, are considered. In order to show the benefits of each smart strategy, a reference case of UCC, where charging starts as soon as an EV user reaches home and stop when the battery is full, is also presented. The focus of this work is only on residential EV charging demand and it is assumed that each EV will charge at its rated power only and charging will be uninterruptible.

3.1 Aggregator-based charging (ABC)

The goal of an EV aggregator is to maximise the profit. Aggregator is not concerned regarding the network constraints or efficiency. It is assumed that aggregator will have an agreement with the EV users such that revenues are fixed, for example, aggregator may have an agreement with the EV users, under which EV users will pay a fixed monthly fee to aggregator. So, effectively the problem of maximising profit boils down to minimise the total charging cost which also supports the interest of EV users. Thus, aggregator indirectly represents the interest of EV users also. The problem for aggregator is to optimise charging starting time for each EV such that total charging cost is minimised while adhering to mobility constraints and requirements of each EV user. The optimisation then takes the following form:
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0032(4)
subject to
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0034(5)
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0036(6)
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0038(7)
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0040(8)
where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0042 is the electricity price at the urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0044 time, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0046 is the charging rate of an EV, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0048, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0050, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0052, and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0054 are arrival time, departure time, charging duration, and optimal charging starting time of the urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0056 EV, respectively, and N is the total number of EVs. Constraints (5) and (6) ensure continuous/uninterruptible charging at a maximum charging rate. Constraints (7) and (8) indicate that both optimal charging start time and charging end time considering optimal charging start time should lie between the expected arrival and departure time of an EV obtained from (2) and (3). Charging duration of any EV can be obtained as follows:
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0058(9)
where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0060 is the power consumption of an EV per 100 km, urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0062 is the daily distance travelled by the n th EV as obtained from (1), and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0064 is the charging efficiency.

3.2 Network operator-based charging (NOBC)

The objective of network operator is to make sure that network is secure, reliable, and efficient. Due to the difference between peak and valley load, there is underutilisation of grid assets and investments. Network operator will attempt to shift the EV load demand to the valley of the load curve so that system's efficiency and asset utilisation is improved. It shown in [12] that minimising losses is the same as minimising the load variance. So, the objective considered here from network operator's perspective is to minimise the load variance. The problem for the network operator is to obtain the optimal charging start time of each EV which minimises urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0066 in conjunction with constraints (5)–(8).
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0068(10)
with
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0070(11)
where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0072 is the base load of the system (without EVs) and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0074 is the mean of the aggregated load demand of T intervals.

3.2 Coordination-based charging (CBC)

The total charging demand of EVs, scheduled by an aggregator, may cause undesirable impacts on the grid as the aggregator does not has any responsibility for stable operation of the system but it primarily aims to maximise its profit. To overcome the conflicting interest of aggregator and network operator, coordination-based charging is taken up. This strategy includes the interest of both aggregator and network operator. In coordination-based charging strategy, both aggregator and network operator will work in coordination to achieve a common objective. The objective considered here is to simultaneously minimise total charging cost and load variance, so as to get both economic and technical advantages. The problem is to optimise charging start time of each EV which minimises urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0076 while satisfying constraints (5)–(8)
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0078(12)
In this paper, the problem of multi-objective is converted into single objective by using the fuzzy method. Using the fuzzy method, the memberships can be obtained. The formula is represented in (13), where urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0080 are values of each objective function during calculation, minimum value of each objective function, and maximum value of each objective function, respectively
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0082(13)
According to this method objective, functions are represented by fuzzy membership functions, which are urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0084 where i = 1, 2. urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0086 and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0088 are the weights assigned to each objective function and can be set by decision makers. Taking the sum of these weighted memberships as a new objective function, the multi-objective problem is converted into a single objective problem. The number of functions considered is two. The new formula is as follows:
urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0090(14)

4 Results and discussion

The EV used in this paper is Toyota RAV4 [14]. Its battery capacity is 21.6 kW h, power consumption per 100 km is 13.9 kW h, and inverter efficiency is 75%. Most of the commercially available EVs in market have onboard chargers belonging to either level 2a (3.3 kW) or level 2b (6.6 kW) charging set [15]. So, in this paper, both 3.3 kW-slow charging and 6.6 kW-fast charging are compared. The scaled residential grid load profile and electricity price of the residential grid, which is used for calculating EV charging cost, are taken from [10]. EVs are schedule one by one. Total time interval is divided into 1440 intervals. Both weights urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0092 and urn:x-wiley:20513305:media:tje2bf00484:tje2bf00484-math-0094 are taken as 1.

4.1 Impact of different charging strategies on the grid load curve

The impact of different EV charging strategies with 3.3 kW charging power considering different penetration levels of EVs on the grid load curve is shown in Figs. 4-1, and the impact on peak-valley difference and load factor is presented in Table 1. In the case of UCC, EV will start charging immediately on arriving home. Therefore, the UCC strategy will exacerbate the already-existing peak which is also shown in Fig. 1. As the penetration of EV increases, the peak load of the system also increases by 12.26, 25, 40.60, and 50.87%. Since EVs charge between 16 and 22 pm, which coincides with the already-existing peak of the system, the peak-valley difference increase and load factor deteriorates as shown in Table 1.

Details are in the caption following the image

Load profiles in the case of CBC (3.3 kW)

Details are in the caption following the image

Load profiles in the case of NOBC (3.3 kW)

Details are in the caption following the image

Load profiles in the case of ABC (3.3 kW)

Details are in the caption following the image

Load profiles in the case of UCC (3.3 kW)

Table 1. Parameters of load profiles under different charging strategies (3.3 kW)
Peak to valley difference, kW Load factor
base load 177.44 54.87
strategy number of EVs
UCC 20 209.07 50.31
40 242.07 46.4
60 282.29 42.25
80 308.69 40.35
ABC 20 174.14 56.61
40 172.74 58.25
60 195.22 54.88
80 261.22 45.41
NOBC 20 159.35 56.61
40 145.11 58.21
60 136.41 59.76
80 127.52 61.29
CBC 20 169.2 56.61
40 156.37 58.25
60 151.11 59.76
80 151.11 61.29

As all EVs are available for charging between 3 and 5 am, when the electricity prices are low, a new peak due to EV charging will arise during this time interval in the case of the ABC strategy as shown in Fig. 2. Although, for low penetration of EVs, the ABC strategy assists in valley filling, but for high penetration, it gives rise to a new peak which will be even larger than the original peak due to simultaneous charging of all EVs. Table 1 also shows that with the penetration of 40 EVs, EVs load demand improves the load curve, but after that the results are opposite. There is an increase in the peak by 8.8 and 34.3% when numbers of EVs are 60 and 80, respectively. In comparison to the UCC and ABC strategies, there is no increase in the peak when the NOBC strategy is adopted as shown in Fig. 3. Due to the minimum load variance objective function, the NOBC strategy tries to fill the valley of the load curve. As the number of EVs increases, more valleys filling and flatness of load curve are achieved as shown in Table 1. To overcome the trade-off between ABC and NOBC strategies, a CBC strategy is proposed. The impact of the CBC strategy on the load curve is shown in Fig. 4. In this case also, there is no increase in peak load and valley filling is also observed. However, the peak-valley difference is greater in comparison to the NOBC strategy and lesser in comparison to the ABC strategy (for any similar number of EVs) as shown in Table 1. This is due to multi-objective function considered for the NOBC strategy. It is also observed that load factor for any particular penetration of EVs is the same in the case of both NOBC and CBC strategies because the total average load (depends on load of EVs) and peak load are the same for both the strategies considering a particular penetration of EVs.

4.2 Charging cost under different strategies

Fig. 5 shows the charging cost of EVs under different strategies. As expected for any particular penetration of EVs, UCC has the highest cost (as EVs charges in evening when prices are more) and ABC has the least charging cost (due to minimum cost objective function). For low penetration of EVs, there is a less difference between the charging cost incurred due to ABC and NOBC strategies, but this difference increases with the increase in the penetration level of EVs. The charging cost in the case of the CBC strategy lies between cost incurred by ABC and NOBC strategies.

Details are in the caption following the image

EVs charging cost under different strategies

4.3 Influence of fast charging on the grid load curve

The peak load increases by 17.17, 36.82, 49.61, and 60.52%, when charging power is increased to 6.6 kW for the UCC strategy as shown in Fig. 6. This shows that the charging rate can have a significant impact on the system load in the case of UCC. Due to high charging rate, the peak-valley difference increases and load factor degrades more for 6.6 kW charging power in comparison to 3.3 kW charging as shown in Table 2. In the case of the ABC strategy, increasing the charging power will generate a new load peak for even less penetration level of EVs as shown in Fig. 7. The percentage increase in the peak load is observed to be 35, 85.2, and 136.37%, respectively, when the number of EVs is increased to 40, 60, and 80, respectively. Fast charging will have more severe effects on the load curve even in comparison to UCC for higher penetration of EVs as confirmed from Table 2. The load profiles under NOBC and CBC strategies with increased charging rate are shown in Figs. 8 and 9, respectively. Charging rate does not cause any significant changes on the load curve if NOBC and CBC charging strategies are adopted.

Details are in the caption following the image

Load profiles in the case of CBC (3.3 kW)

Details are in the caption following the image

Load profiles in the case of NOBC (6.6 kW)

Details are in the caption following the image

Load profiles in the case of ABC (6.6 kW)

Details are in the caption following the image

Load profiles in the case of UCC (6.6 kW)

Table 2. Parameters of load profiles under different charging strategies (6.6 kW)
Peak to valley difference, kW Load factor
base load 177.44 54.87
strategy number of EVs
UCC 20 221.76 48.17
40 272.44 42.34
60 305.44 39.67
80 333.58 37.89
ABC 20 177.44 56.61
40 265.29 43.23
60 397.29 31.97
80 529.29 25.63
NOBC 20 161.29 56.61
40 147.62 58.24
60 137.17 59.75
80 127.94 61.28
CBC 20 171.52 56.61
40 159.83 58.24
60 151.11 59.75
80 151.11 61.28

4.4. Influence of charging efficiency on charging cost

Charging efficiency is defined as the percentage of power drawn from the grid that is actually taken up by a vehicle battery [16]. There is always some loss of electricity while charging EVs due to AC to DC conversion. The impact of increasing efficiency on the charging cost of EVs is shown in Table 3. The charging cost can be improved by increasing the charging efficiency for all kinds of EV charging strategies.

Table 3. Influence of battery charging efficiencies on charging cost
Charging strategy
Efficiency UCC ABC NOBC CBC
0.75 140.42 72.49 78.58 75.85
0.85 125.08 63.74 68.79 66.69
0.95 112.96 56.98 61.09 59.62

5 Conclusions

The aim of this work is to compare smart EV charging strategies from three different stakeholders' perception. The results show that the charging strategy considering only aggregator's interest (ABC strategy) will give a minimum charging cost and assists in improving the load profile but only for a low penetration level of EVs. Fast charging along with high penetration of EVs under the ABC strategy will cause adverse effects on the load curve. The adverse effects will be even more precarious than with UCC. Both, for slow and fast charging with the increase in penetration of EVs, more valley filling and load curve flattening are achieved for the NOBC strategy. Simulation results fall in favour of both aggregator and network operator's interests in the CBC strategy as the multi-objective function is considered. It not only helps to achieve economic cost, but also assists in valley filling. Thus in future, a multi-objective function as taken up in this paper can be adopted for coordinate charging so as to provide benefit to all stakeholders. It is also found that upon increasing the charging efficiency, the charging cost for all kinds of EV charging strategies reduces.