Charging Stations

Charging/discharging of PEVs takes place at residential or public electrical installations, each equipped with a number of power connections serving one to many PEVs at the same time.

An installation which is able to do both supplying energy to exactly one PEV charging its battery and accepting feedback of energy from the same PEV discharging its battery is termed charging station (CS) in this study. The most relevant characteristic of a CS is the maximum power it is limited to. Typical power connections which are used in this case study are defined in Table 4.7.

Table 4.7: Typical power connections for PEV charging/discharging.

Power [kW]

Voltage [V]

Max. current [A]

Type Charging/discharging time for 20 kWh^a)

3.6 230 16 AC (single phase) 5.5 h

7.2 230 32 AC (single phase) 2.8 h

11 400 16 AC (three phases) 1.8 h

22 400 32 AC (three phases) 1 h

43 400 63 AC (three phases) 30 min

120 400 300 DC 10 min

a) Charging/discharging times are calculated using theconstant current scheme only.

Realistically applying the the constant voltage scheme at an SOC above 80 % extends this duration by a factor depending on the specific battery [5].

Besides the maximum power a CS can supply or accept, the temporal distribution of the power values at all times during a stop is an influencing factor when assessing the power system impact. The followingscheduling strategies are therefore investigated:

• Dumb charging

Immediately charging a battery once it is connected to a CS is termeddumb charging in the remainder of this work. No information regarding the charging duration is required allowing to charge the battery as fast as possible. Battery aging costs and the impact on the power system are thereby neglected. Besides, the possibility to exploit variable electricity prices is spared.

• Mean charging

With themean charging strategy a battery is evenly charged with the same power over the entire charging period. This strategy requires knowledge of the charging duration and the battery’s target SOCprevious to the beginning of charging. Battery aging costs or the impact on the power grid are as little considered as variable electricity prices.

• Price-responsive charging

The scheduling approach presented in Section 2.4 implements aprice-responsive charging strategy by adjusting charging power based on temporally resolved price data to minimize charging costs. Besides electricity prices, this strategy again requires knowledge of the charging duration and the battery’s targetSOC previous to the beginning of charging to yield optimal results. Battery aging costs are inherently considered by the optimization algorithm resulting in a cost-minimal charging schedule. To reduce negative impact on the power system’s load curve, prices reflecting the system’s state have to be defined to allow for lower prices in case of power excess and higher prices in the opposite case.

When there are multiple periods with an equal cost-minimization potential, the power demand is equally distributed over all periods equivalent to the mean charging strategy.

This is only possible because calendar aging is not considered in the implemented battery model. Otherwise, at each period the SOC would have an additional influence on the temporal charging decision as described in [35].

• Price-responsive charging/discharging

Besides price-responsive charging the scheduling approach presented in Section 2.4 also provides the possibility for discharging. This strategy is termed price-responsive charg-ing/discharging. The same requirements regarding knowledge of electricity prices, the charging/discharging duration, and the battery’s targetSOCprevious to the beginning of charging/discharging apply. As with the price-responsive charging strategy, this strategy can prevent increasing peaks in the regular demand. To some extent it additionally may even decrease those peaks at some points in time at the cost of a demand increase at others.

In the Singapore PNM, there are 117 852 different locations where CSs can be placed, each one in principle allowing for a different maximum power. The proper placement strategy along with the determination of the maximum power is a research area of its own and is discussed on a high level in Section 4.3.4.3. In this study, CSs are installed at all locations. The number of the installed stations at a location equals the maximum number of simultaneously charging/discharging PEVs at this location. This way each PEV can connect to a CS whenever its itinerary allows doing so. Initially, each location is equally equipped with a fast CS supplying

a maximum power of 120 kW. Other power connections defined in Table 4.7 are investigated in the context of the sensitivity analysis in Section 4.3.3.

4.3.1.2 Scheduling of Battery Energy Storage

The scheduling approach for battery energy storage is applied for both demand-responsive charging and energy back-feeding. The dynamic programming algorithm computes a globally optimal solution sufficiently fast for real-time operation of a single PEV. Computing optimal charging/discharging schedules for a PEV population of an entire city, however, requires decreasing computational complexity. Prices are therefore considered ex post and are thus not updated once made available. This way continuous re-calculation after each period of time using the rolling horizon approach can be avoided. Instead, the rolling horizon approach is applied for each stop thereby calculating the profit-maximizing charging/discharging schedule for the entire lookahead horizon while only taking the time periods until the next trip start into account. Further parameters include the duration of a time period for which prices and charging/discharging power are fixed to 15 minutes, a lookahead of 24 hours, a battery capacity of 20 kWh with an initialSOCof 1 and anSOCdiscretization of 1 %, as well as equal charging and discharging efficiency each of 0.89. In case of demand-responsive charging the minimum power is set to 0 kW, disallowing energy back-feeding. Dropping this restriction and thereby additionally allowing for demand-responsive discharging, the minimum power is set to the negative maximum power of the connected station.

For the price-responsive scheduling strategies electricity prices are set to follow the temporal power demand distribution meaning they have their minimum at the time of the absolute demand valley while the maximum price is only valid for the time period of peak demand.

Prices are not calculated only once but instead they are continuously re-calculated after the sequential processing of each trip. Calculations are based on the aggregation of the regular and the additional power demand of the already scheduled agents including the last one but leaving out trips which have not yet been scheduled. This way, the absolute daily demand valley is continuously shifted in time whenever a trip is scheduled allowing for evenly filling the load area around the valley instead of creating a new peak at that very time. Electricity prices are calculated according to

p(t) = 1− PD,max−PD(t) P_D,max−P_D,min

!

·(pmax−pmin) +pmin (4.2) having a minimum power demandP_D,min of 4 433 MW, a maximum power demandP_D,max of 6 340 MW, a minimum pricepminof 0.135 S$/kWh², and a maximum pricepmaxof 0.359 S$/kWh.

P_D,min andP_D,max result from Figure 4.2,pmin is arbitrarily set to half the consumerelectricity tariff of Singapore on the day on which the power demand used for planning the Singapore PNM occurred, and p_max is set to the value for which the average of all electricity prices equals the consumer electricity tariff of Singapore of 0.27 S$/kWh. The impact of price variations are discussed in Section 4.3.3. The remuneration for providing energy to the grid equals the electricity price. This way, at times of high power demand, the electricity price is

2 S$ 1 equals 0.69 US$ (19th December 2016).

high, preventing the PEV to charge, while at the same time the remuneration is also high, incentivizing providing energy to the grid.