A Study of Smart Grid Effects on Electric Vehicle Management Considering the Change of the Power Capacity Mix

2.1. Charging Capacities of EVs

The charging capacity depends on the specifications of the charging devices. The specifications of charging devices can be classified with relatively simple speculations. The USA Department of Energy groups charging devices into 1.44, 3.3 and 60-150 kW. The categories used by EDF, the French Electricity Provider, are 3, 6, 24, 43, and 150 kW. The larger the capacity of the charger, the faster it can charge [31].

The advantage of rapid charging is the short time period required for complete charging. However, it is not generally used because of the associated high cost and potential shortening of the battery life-span. Rheinisch-Westfӓlische Technische Hochschule Aachen (RWTH) [32] forecasted that 90% of EVs would be charged by slow charging and the remaining 10% by fast charging. Based on this projection, the IEA [18] has assumed that the average charging capacity would be 7.3 kW, given that the slow charging capacity is 3.7 kW and the fast charging capacity is 40 kW [18, 32]. This study has implemented these assumptions.

2.2. Discharging Capacities of EVs

EVs transfer the surplus electricity to the power grid through discharging. Kempton and Tomić [12] defined the EV energy transfer to the power network using equation (1).

(1)

where P_VEH (kW) is the maximum electric power transferable through V2G, E_S (kWh) is the battery-stored energy available as direct current transferable to an inverter, d_D (km) is the driving distance when the battery of an EV is fully charged, d_RB (km) is the minimum range buffer driving distance required for a driver, η_PEV/PHEV (kWh/km) is the average driving efficiency of pure EVs (PEVs) and plug-in EVs (PHEVs), η_INV is the efficiency of the inverter where the direct current of the battery converts into alternating current, and, finally, t_DISP (hour) is the time that the stored energy in EVs is dispatched.

In accordance with the IEA [18] and SNE Research [33], E_S is assumed to be 30 kWh for PEVs and 10.7 kWh for PHEVs. η_PEV/PHEV is assumed to be 0.15 kWh/km through a review of Korean cases. d_D is influenced by various features such as driving behavior, vehicle type, and the charging schedule of drivers. In South Korea, the average daily driving distance is 28 km per driver.

d_RB is not an engineering figure that can be calculated through a study of car features. d_RB is determined by the driver or the car operating company. The buffer driving distance is reserved for unexpected journeys to locations such as convenience stores or hospitals. Through interviews with drivers in California, USA, Kurani et al. [34] suggested that 32 km, which is around 63% of the average daily driving distance per person in the USA, was the range buffer recognized by most drivers. Based on these findings, the range buffer driving distance in this study was set as 20 km, which is 70% of the average daily driving distance per driver in South Korea. As PHEVs can be operated with their internal combustion engine even when they are discharged entirely, d_RB of a PHEV was set as zero. η_INV was set as 0.93, conforming to the assumptions made by Kempton and Tomić [17].

t_DISP can range between a few minutes for spinning reserve and a few hours for peak load response. According to the IEA [18], peak load occurs for three to five hours per day. As this study examines the effects of EVs on peak load, t_DISP was set as three to four hours. The forecasted number of disseminated PEVs and PHEVs are similar in agreement with the IEA [35]. Therefore, P_VEH (kW) operates as an average of the PEVs and PHEVs. Given the assumptions above, the discharging capacity was calculated to be about 4 kW.

2.3. The Controlling Effect of SGs on the EV Impact on Peak Load

The proportion of EVs that are connected to the power grid during peak load greatly affects the overall impact of EVs on the load of the electric power network. The IEA [18] has assumed that the proportion of EVs connected to the power grid during peak hours, without SG utilization, is 50%. Korea Power Exchange (KPX) [36] assumed the proportion is 30% and Choi et al. [19] assumed 40%. This study considers all precedent studies on the electric vehicle charging share during peak hours.

The SGs can lessen the load by partially disconnecting EVs from the power grid or lowering the charging capacity during peak load. The EV charging management market is expected to grow with the already matured technologies. As V2G is cost intensive and demands additional infrastructure to enable bilateral power flow between EVs and the power grid, however, it is forecasted to grow at a slower rate than the EV charging management market. The battery of an EV participating in V2G experiences functional degradation earlier than that of a non-participating EV. The costs of the functional degradation of the batteries must be incorporated into the compensation offered to the consumer by the power provider. However, it is not easy to quantify functional degradation into a monetary value. Furthermore, as a consumer participates in V2G, the shortened driving distance necessitates frequent charging, thereby resulting in lower acceptance of V2G over the grid to vehicle (G2V).

Table 1 provides three types of SG utilization in parallel with the IEA [18]. Without SG utilization (labeled SG Non-Utilization), the proportion of G2V during peak load time is 30% or 40% or 50% with no V2G. The two other types, in accordance with dynamic pricing and the degree of SG utilization, are divided into SG Low-Utilization and SG High-Utilization.

Table 1 shows that a total of 9 analysis cases have been drawn based on the degree of SG utilization and the charging and discharging share during the peak hours; f_EV (G2V) is decreased by 10%p, f_EV (V2G) is increased by 5%p as the degree of SG utilization rises. The following is an overview of the cases.

• Case 1: f_EV (G2V) is 30%. There is no SG utilization.
• Case 2: f_EV (G2V) is 10%p less than Case 1 and f_EV (V2G) is 5%p more than Case 1.
• Case 3: f_EV (G2V) is 20%p less than Case 1 and f_EV (V2G) is 10%p more than Case 1.
• Case 4: f_EV (G2V) is 40%. There is no SG utilization.
• Case 5: f_EV (G2V) is 10%p less than Case 4 and f_EV (V2G) is 5%p more than Case 4.
• Case 6: f_EV (G2V) is 20%p less than Case 4 and f_EV (V2G) is 10%p more than Case 4.
• Case 7: f_EV (G2V) is 50%. There is no SG utilization.
• Case 8: f_EV (G2V) is 10%p less than Case 7 and f_EV (V2G) is 5%p more than Case 7.
• Case 9: f_EV (G2V) is 20%p less than Case 7 and f_EV (V2G) is 10%p more than Case 7.

The effect of EVs on the peak load can be expressed as equation (2) [18]. The coefficient δ is determined by multiplying the proportion of EVs charging during peak load time by the average charging electric power (7.3 kW) of each EV. The coefficient ε is determined by multiplying the proportion of EVs discharging during peak load time by the average discharging electric power (4 kW) of each EV. N_PEV/PHEV is the number of EVs at time (t). Taking advanced metering infrastructure (AMI) into account is important because V2G is available after the environment is prepared where demand response is possible through AMI.

(2)

There is a time-lag between the establishment of AMI and the functioning of V2G. V2G must first be commercialized with V2G-supported AMI being deployed in advance. South Korea plans to distribute smart meters to every household by 2020 and optimize the demand response system for reasonable energy consumption through additional updates of the metering system by 2030. It may be assumed that AMI deployment would be complete by 2030. However, the effect of AMI on V2G will differ according to the timing of V2G commercialization and the degree of V2G-supported AMI dissemination. This study has assumed that the V2G-supported AMI roll-out will begin in 2020 and be completed in 2035, with V2G being available as soon as V2G-supported AMI is established.

It is assumed that the V2G-supported AMI deployment will show exponential growth in the early stage, gradually slowing down and finally maturing when AMI is distributed to almost all households. The proportion of the V2G-supported AMI roll-out in a particular year between 2020 and 2035 can be formulated as equation (3) [18].

(3)

The coefficient r is a correction coefficient to match the maximum AMI deployment to the target year, and the STEP function shows that the proportion of AMI roll-out keeps increasing until the target year for the entire distribution of AMI. T_Target is the year during which V2G-supported AMI is completely deployed nationwide, and T_Now is the year when V2G-supported AMI is built. The STEP function can be defined using equation (4).

(4)

Choi et al. [19], using a conjoint analysis, forecasted the number of EVs that would be disseminated by 2035 in a scenario where governmental support including tax deduction continues and the development of the related technologies progresses. By incorporating their research into the present situation in Korea, the number of EVs was estimated at about 10.3 thousand in 2015 and 1,775 thousand in 2035 (Table 2).

The peak load growth caused by EVs based on the assumed forecast has been calculated and is shown in Fig. (2). Without SG utilization, the peak load increase caused by EVs is 3,888 MW for the year 2035 in Case 1. The peak load is increased by 5,183 MW in Case 4 and by 6,479 MW in Case 7 in 2035. However, in the cases of high SG utilization such as Case 3, Case 6, and Case 9, the peak load increments remain at 586 MW, 1,882 MW, and 3,187 MW, respectively. These amounts are roughly 3,302 MW less than the peak load increment when SGs are not utilized (Fig. 2).

3.1. WASP Model and Major Assumptions

The Wien Automatic System Planning (WASP) model was originally developed in 1972 for the IAEA by the Tennessee Valley Authority and the Oak Ridge National Laboratory in the USA to analyze the economic competitiveness of nuclear power in comparison to other sources of electricity generation in meeting electricity requirements of a country or region in the medium- to long-term period [18]. WASP is continuously upgraded by the IAEA and is currently available as version 4. WASP determines the generating system expansion plan that meets demand at minimum cost while satisfying certain user-specified constraints for the system. The objective function (B_j) of the optimization problem in WASP is expressed in equation (5) [37].

(5)

where j is the alternative expansion path (program), t is the year of the study period (t = 1,2,...,T), T is the length of the study period, I is the investment (capital) costs, S is the salvage value, F is the fuel cost, OM represents the operation and maintenance costs, and U is the unserved energy cost, with all costs being discounted to a reference year at the annual discounting rate.

The input data for the WASP model are load forecast, discount rate, Loss Of Load Probability (LOLP) as a constraint, and technical and economical characteristics of existing and candidate power plants such as investment cost, fuel cost, OM costs, heat rate, and Forced Outage Rate (FOR). The output data of the WASP model are build schedule, generation, fuel consumption, costs, and emissions.

The target year that this study analyzed is 2035 and nine cases shown in Table 1 were examined. Additionally, the baseline case in which the deployment of EVs and the effect of SGs are not reflected was also examined.

To understand the electric power demand, the demand forecast in the 2^nd Korean Energy Master Plan (2014-2035), the forecast of EV electricity consumption (GWh) [19], and the forecast of peak load (MW) in accordance with EV deployment and SG application were applied. The electricity consumption in Cases 1~9 was assumed to be similar. This was because the SG does not reduce the demand for the use of EVs although it does play a role in reducing or migrating peak load increases by EV charging. As EV charging costs become lower, the demand for EV use can be enhanced. This study applied only the changes of peak load caused by SGs.

In addition to the demand for electricity, the remaining input data used for the WASP model were provided by the 2015 7^th Basic Plan for Long-term Electricity Supply and Demand (2015-2029) [38]. The WASP model was set to produce a cost-minimizing plan that satisfies the system reliability standard of the LOLP 0.3 day/year and a reserve margin of around 22%. The discount rate was set at 5.5%. These criteria were derived from the 7^th Basic Plan for Long-term Electricity Supply and Demand (Table 3).

The 2035 capacities of some power sources were preliminarily fixed. In the analysis fixing the nuclear power capacity, 29% of total capacity in the baseline case was used as the nuclear power capacity in all cases because the 2nd Korea Energy Master Plan, which was made in January 2014, set the percentage of nuclear power plants in 2035 as 29%. The percentage of renewable energy is in agreement with the figure provided by the 2^nd Korea Energy Master Plan (Table 4) [39].

There are some reasons why power source shares must be preliminarily reflected. Nuclear power has the most limitations in terms of power plant construction and the associated huge construction costs, long construction periods, and difficulty in securing a building site. The number of nuclear power plants estimated using WASP can be much higher than that of actual constructions because of the low unit cost for power generation. In addition, the construction period of a nuclear power plant is more than 10 years and the plants currently planned or under construction can be incorporated by 2020. It is important that these aspects are reflected. As for renewable energy sources, it is necessary to determine the preliminary forecasts of the installed capacity and generated electric power of renewable energy and then make a residual load duration curve deducting the renewable energy sources.

3.2. The Changes in Generation Mix (GWh) Caused by EV Deployment and the Analysis of the Effect of SGs

According to the analysis, the baseline case showed that the total installed capacity is 147,516 MW, as presented in Table 5. The capacity share of nuclear power plants and natural gas-powered plants is 29%, and the capacity share of coal powered plants is 32.6%. The installed capacity of Case 1 reflects the 2035 scenario where EVs are disseminated by government support, but without the effect of SGs. In Case 1, nuclear power capacity was also fixed as 42,829MW, which is 29% of the share in the baseline case. As EVs cause the peak load to increase, the total installed capacity is expanded to 151,516 MW, and the share of nuclear power in terms of installed capacity decreased to 28.3%. The capacity by power supply from Case 1 was applied to Cases 2 to 9, not reflecting the effect of SGs on the changes of the installed capacities. This is because, by fixing the installed capacity mix (MW), the changes in the power generation mix (GWh) due to SG utilization can be examined, in addition to the generation costs and CO₂ emissions.

Table 6 shows how the generation mix varies in the nine cases in addition to the baseline case. The total power generation is 756,963 GWh in the baseline case and increases to 874,516 GWh in Case 1. This result is attributed to the increase in electric power demand as EVs are disseminated. Overall, the shares of coal and natural gas powered generation are increased in Cases 1 to 9 compared to the baseline case. However, all the nuclear power generation in Cases 1 to 9 is fixed as 37% due to the similar total demand and fixed installed capacity regardless of the degree of SG utilization. Cases 1 to 3 show that with a higher degree of SG utilization, more coal is produced, but less natural gas is produced. The same result is derived for Cases 4 to 6 and 7 to 9. By applying the effects of SGs, the peak load lessens and the use of natural gas power plants as a power generation unit for peak load becomes lower. The coal powered generation, in contrast, increased to replace the decreased natural gas-powered generation due to the SG utilization. This phenomenon occurs strongly in Cases 3, 6, and 9, where the degree of SG utilization is the highest.

Table 7 below illustrates the average unit cost and CO₂ emission by analyzed cases. The average unit cost was calculated by multiplying power generation by power source from the WASP analysis by the unit cost, adding them, and dividing the calculated number by total generation. The settlement price in 2014 was premised as the unit cost by power source. Some details of the settlement price (won/kWh) of major power supply are as follows: nuclear 54.96, coal 65.79, oil 221.32, and LNG 162.34. The greenhouse gas emission was drawn through the WASP-IV model.

In Case 1, the average unit cost of power generation increases by about 2.9 won/kWh compared to the baseline case. However, in Cases 2 and 3 that reflect the SG effect, it decreases slightly. Meanwhile, the CO₂ emissions are higher when the SG effect is expressed. Cases 4~6 and Cases 7~9 are also similar. It is because the share of natural gas-powered generation decreases and that of coal increases. As long as the proportions of nuclear power and renewable energy remain limited, the peak load reduction or migration has the effect of increasing the CO₂ emissions in the power generation sector.

3.3. The Changes in Installed Capacity Mix (MW) and the Effect of SGs