A rule-based energy management scheme for

Renewable Energy 125 (2018) 384e400

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

A rule-based energy management scheme for uninterrupted electric vehicles charging at constant price using photovoltaic-grid system Abdul Rauf Bhatti a, b, Zainal Salam a, * a b

Department of Electrical Engineering, Government College University Faisalabad, Faisalabad 38000, Pakistan Centre of Electrical Energy Systems, Universiti Teknologi Malaysia (UTM), 81310 Johor Bahru, Johor, Malaysia

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 28 February 2018

This work proposes a rule-based energy management scheme (REMS) for electric vehicle (EV) charging from photovoltaic-grid (PV-grid) system. The main feature of this scheme is that it provides uninterrupted daytime charging at a constant price. In order to simulate the system, the models of PV output power, EV power demand, state of charge (SOC) estimation of energy storage unit (ESU) and grid electricity prices are developed. The uninterrupted and constant price charging is achieved by managing the energy flow between PV, ESU and grid according to the rules defined by REMS. Furthermore, the valleyfilling operation is implemented during the grid off-peak hours. The resiliency of REMS is validated under various weather conditions, different ESU prices and at grid parities. For comparison, its performance is benchmarked against the standard grid-based EV charging. The results demonstrate a decline in charging price by 16.1% besides reducing the burden on the grid by 93.7% with the implementation of REMS. In addition, the vehicle-to-grid (V2G) technology is incorporated in the charging system to improve the payback schedule of the existing PV-grid system. © 2018 Elsevier Ltd. All rights reserved.

Keywords: Constant price charging Renewable energy Electric vehicles Energy management scheme Photovoltaic-grid system Rule-based algorithm

1Introduction The direct charging of electric vehicles (EV) using standard electrical grid outlet is widely accepted due to its simplicity and the provision of unlimited energy source [1]. However, it imposes an extra burden on the supplydparticularly during the daytime where the electricity demand is at its peak. The grid overloading results in voltage/frequency deviation, distribution losses, and degradation in the power quality [2]. To overcome these consequences, a hybrid charger using renewables energy (RE) is proposed [3,4]. Additionally, the environmental benefits of using RE source are well understood among power engineering practitioners for providing electrical energy [5e7]. Among the RE sources, solar photovoltaic (PV) appears to be well-suited for daytime charging, where the demand is highest and coincides with the peak tariff [8,9]. Photovoltaic-grid (PV-grid) systems have been spread in many countries because of its potential long-term benefits [10,11]. Furthermore, the PV system can be installed in existing buildings or

* Corresponding author. Centre of Electrical Energy Systems, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310, Johor Bahru, Malaysia. E-mail addresses: [email protected], [email protected] (A.R. Bhatti), [email protected], [email protected] (Z. Salam). https://doi.org/10.1016/j.renene.2018.02.126 0960-1481/© 2018 Elsevier Ltd. All rights reserved.

car parking space [12]. In the latter case, the roofed structure provides shelters from sun and rain, which is favorable in hot climate countries. In addition, the availability of solar power enables the prospect for “charging while parking” [13], which is ideal for shortterm charging in shopping malls and office parking lots. Despite these advantages, the operation of the hybrid PV-grid EV charging (henceforth called as PV-grid) is quite complex due to the intermittent PV power, fluctuating grid electricity prices and the random nature of EV power (for charging) demands. Thus, PV requires the adoption of an energy storage unit (ESU) to compensate fluctuations and to meet the energy demand during low PV power generation [14]. However, if ESU is included in PV-grid system, its state of charge (SOC) must be maintained to ensure battery longevity [15]. Inevitably, an efficient energy management scheme is required to optimize the utilization of these hybrid power sources with the specific objectives required by the EV charging station owner as well as the users. The energy management schemes for EV charging systems are mainly based on the optimization or rule-based algorithms [16,17]. The former minimizes a certain objective function that represents the energy flow among the sources and load. It also adopts specific financial model regarding profitability and losses [18,19]. The commonly used optimizers are the genetic algorithm (GA), particle

A.R. Bhatti, Z. Salam / Renewable Energy 125 (2018) 384e400

swarm optimization (PSO), linear/nonlinear programming (LP/ NLP), dynamic programming (DP) and the stochastic dynamic programming (SDP) [16]. On the other hand, the rule-based algorithms rely on human knowledge and experiences of the system to create a set of rules for efficient interaction between the energy sources and load (EV) [20]. They are preferable for real-time energy management, as they are computationally efficient and are able to give an exact solution to the desired output conditions [21,22]. Normally, the major objectives of the existing energy management schemes are to: 1) reduce the grid burden during load peak hours [23e28], 2) provide EV charging at possible low price [23,25,28e30] and 3) increase the profit of charging station [23,29,31,32]. These objectives are generally approached by: 1) shifting the time of charging to the grid off-peak hours (i.e. valley filling) [24,28,33] and/or 2) imposing the vehicle-to-vehicle (V2V) and vehicle-to-grid (V2G) operations (peak-shaving) [23,30,34]. These approaches create unexpected interruptions in the charging process as it is terminated whenever the transition from off-peak to peak load takes place [24]. This reduces the autonomy of charging the EV freely and independently. More so, under the existing schemes, the charging prices are varied based on: 1) present realtime energy price of the source being utilized [35,36], 2) users' preferences [23,32], 3) V2G/V2V operations [23] or 4) shifting the charging to the time of minimum electricity price [28,29,37,38]. Thus, by adopting these schemes, a fixed/constant charging price cannot be achieved. Consequently, the unexpected interruptions in the process, as well as the uncertainties in charging price, create inconvenience for the EV users. These problems are not faced by the users of internal combustion engine (ICE) vehicles during the time of fuels (petrol/diesel) filling. Based on these motivations, a rule-based energy management (REMS) scheme for the PV-grid charging station is proposed. It incorporates the following features, which are also considered as the major contributions of this work: 1) charging EV at a constant price (cents/kWh), similar to the fuel filling for ICE vehicles (cents/liter). 2) providing an uninterrupted charging during the daytime without imposing V2V/V2G operations. 3) reducing charging price as well as grid burden compared to those under standard grid charging process. The proposed REMS is developed for the office-based charging stations located in the areas of solar grid parity. Another important aspect of the system is the charging cost which is maintained lower than average grid electricity price. 2Modeling of PV-grid charging system Following the previous trend, the design and evaluation of the charging station are based on simulation of interconnected various component models [3]. The electrical power and control architecture of the system is shown in Fig. 1 [26,27]. All the system components are linked via a common dc bus. The PV array is connected to a dc-dc converter, with the maximum power point tracking (MPPT) control. The utility grid is connected through a bidirectional inverter, which is responsible for performing the power conversion between dc bus and grid. The ESU connects to the dc bus using a bidirectional dc charger [32]. The unidirectional dc charger regulates the charging power and provides a decentralized control to the EV user [26]. This charger communicates with the central controller to transmit the information of EV power demand. Then the central controller instructs the power converters of energy sources, i.e. PV array, ESU and utility grid to control the energy flow. The REMS is embedded

385

Fig. 1. Architecture of PV-grid EV charging system.

in this controller to meet the EV power demand in real-time. 2.1. The EV power demand model The EV power demand (EV_Dmd) model is based on the work in Ref. [39]. The power demanded by a single EV at time t is calculated as

Pev;t ¼ Pev;req St wt ct

(1)

where Pev,req is the power required by the EV battery to increase its state of charge (SOC) from the initial (SOC0) to maximum (SOCmax) value in time step Dt. Variable St, is the control signal to check the SOC, while wt, and ct represent the working days and office hours, respectively. For multiple connected EVs, the Pev,t is summed for the total EV_Dmd, i.e.

EV Dmdt ¼

8 N PV_Pwr, the PV alone cannot satisfy the EV_Dmd; the remaining is fulfilled by ESU or grid. If the grid is at off-peak conditions, the Gd2EV is activated. At the same time, the SOC is checked; if SOC < SOCU, the ESU is also charged by the grid (Gd2ESU) at GE_Pr using Mode 6. Thus, both EV and ESU are taking benefit of low GE_Pr by means of valley-filling operation. On the other hands, if the grid is loaded and the Avl_ESU_Pwr EV_Dmd, the demand is fulfilled by ESU alone using ESU2EV. However, if Avl_ESU_Pwr < remaining EV_Dmd, the grid helps ESU to provide the EV need (Gd2EV) using Mode 3. On the extreme case, if both PV and ESU have insufficient energy, the remaining EV_Dmd will be met by the grid (Gd2EV) using Mode 3. It is worth mentioning that the energy is sold to EV at constant Chrg_Pr irrespective of the source. Moreover, the V2V and V2G operations are not imposed on the user in the operation of charging station even during the overload scenario; hence providing autonomous charging.

3.6. Mode 6: grid to ESU or valley-filling (Gd2ESU) Mode Gd2ESU is activated when the grid is at off-peak and PV_Pwr is not sufficient to increase the SOC to SOCU. The process of charging batteries during grid off-peak is called valley-filling. The amount of energy transfer varies based on two situations. In Situation 1, if a surplus PV_Pwr is available, the ESU is charged from PV on a priority basis; the remaining Req_ESU_Pwr is taken from the grid, i.e.

Gd2ESU ENR ¼ fReq ESU Pwr ðPV Pwr EV DmdÞg Dt (23) In Situation 2, the entire Req_ESU_Pwr is taken from the grid if no surplus PV_Pwr available. This is expressed as

4.2. Under load scenario Fig. 3 shows the operation during the under-load scenario. In this case, the generated PV_Pwr EV_Dmd; thus, the surplus PV_Pwr can be fed to ESU to increase its SOC. Alternatively, it can be sold to the grid for financial gain. The extra PV_Pwr is prioritized to charge ESU (PV2ESU) using Mode 4. However, if Req_ESU_Pwr > surplus PV_Pwr, the remaining demand is fulfilled by a valley-filling operation using Mode 6 (Gd2ESU). On the other hand, if the surplus PV_Pwr > Req_ESU_Pwr, the remaining energy is sold to the grid (PV2Gd) using Mode 5. However, if ESU is already fully charged, then the entire surplus PV_Pwr is sold to the grid.


389

Overload Scenario EV_Dmd > PV_Pwr

Chrg_Pr = PV_Pr 1.0 Mode 1 PV2EV @ Chrg_Pr EV_Dmd NotFull valley-filling No

No

SOC > SOCL No

Yes

GE_Pr < Chrg_Pr

Mode 3 Gd2EV @ Chrg_Pr EV_Dmd Full

Yes

(EV_Dmd PV_Pwr) > Avl_ESU_Pwr

Mode 3 Gd2EV @ GE_Pr EV_Dmd Full Mode 2 ESU2EV @ Chrg_Pr EV_Dmd Full SOC < SOCU

Yes

Mode 2 ESU2EV @ Chrg_pr EV_Dmd NotFull SOC = SOCL

No

SOC < SOCU

Yes

Mode 6 Gd2ESU @ GE_Pr SOC = SOCU

Mode 3 Gd2EV @ Chrg_Pr EV_Dmd Full

Fig. 2. Operation during the overload scenario.

4.3. No load scenario During no-load, the total PV_Pwr has the option 1) to charge the ESU and/or 2) to feed the grid. The SOC of ESU is checked first; if it is less than the SOCU limit, the PV_Pwr is utilized to charge the ESU (PV2ESU) using Mode 4. However, if the PV_Pwr exceeds the Req_ESU_Pwr, the excess power is sold to the grid (PV2Gd) using Mode 4 as shown in Fig. 4. On the other hands, if Req_ESU_Pwr > PV_Pwr, the ESU can get the remaining power from the grid (Gd2ESU) through a valley-filling (Mode 6). However, if the ESU reaches SOCU limit, the entire PV_Pwr is sold to the grid (PV2Gd). 4.4. Idle scenario The REMS based operation of charging station during the idle scenario is shown in Fig. 5 The charging of ESU takes place if SOC < SOCU and the grid is at off-peak. The overall operation during the idle scenario is accomplished through single operating mode i.e. Mode 6. The idle condition plays a very important role in reducing the economic losses of charging station. Nevertheless, to ensure the lossless operation of the system while charging a specific number of EVs, it is important to find the minimum size of PV and ESU. In this work, the sizes are calculated by optimizing the financial model of the system. 5Financial model for system sizing Resource optimization is a major factor in the assessment of the

effectiveness of renewable energy systems [73]. The optimal components design for grid-connected photovoltaic systems should be determined with consideration of system operation [74]. This is because the reliance of future energy demand on the system is based on its payback period and particular electrical grid parity prices [75]. In this work, the size of the charging station's infrastructure is calculated based on LCOE of PV as well as ESU. Thus, the return on investment (ROI) i.e. the payback schedule is automatically taken into account during the calculation process. To provide a PV-grid EV charging facility that guarantees an interruption-less daytime charging at a constant price (PV_Pr d 1.0 cents/kWh), the first step is to determine the minimum size of the charging station to avoid annual economic losses while achieving the objectives of REMS. For the size calculation, the optimization process utilizes control strategies (REMS in this research) that define objective functions and constraints [76]. The REMS tracks and directs the energy flow among various components of the system by means of the power converters switching. Besides that, it also records the respective prices of the energy purchasing and selling. The price of the purchased energy (in cents/kWh) can be computed as [32];

t Peng

8 < ðPV2EV ENR þ PV2ESU ENR þ PV2Gd ENRÞt PV Pr ¼ þðESU2EV ENRÞt ESU Pr : þðGd2EV ENR þ Gd2ESU ENRÞt GE Prt (25)

In (25), the purchasing price is the sum of the prices of energy purchased from PV, ESU, and the grid. The factor (PV2EV_ENR þ PV2ESU_ENR þ PV2Gd_ENR) represents the energy

390


Underload Scenario EV_Dmd > 0 AND EV_Dmd 0

Mode 5 PV2Gd @ GE_Pr– 0.1

SOC < SOCU No

Yes

Mode 4 PV2ESU @ PV_Pr SOC < SOCU No

Yes

PV_Pwr > Req_ESU_Pwr

GE_Pr < Chrg_Pr

Mode 4 PV2ESU @ PV_Pr SOC = SOCU Yes


Fig. 4. Operation during the no load scenario.

Mode 5 PV2Gd @ GE_Pr– 0.1 valley-filling

No


Idle Scenario PV_Pwr = 0 AND EV_Dmd = 0

No

SOC < SOCU

Yes

GE_Pr < Chrg_Pr

valley-filling

No

Yes


391

To ensure the resiliency of the REMS in providing uninterrupted charging at a constant price under different number of office working days, the system sizing is done for various cases of office non-working days throughout the year. For this purpose, the PSO is applied to four cases of holidays namely: i) all holidays that is by considering all public holidays and two weekend days, namely Saturday and Sunday, ii) one weekend that is all public holidays and Sundays, iii) no weekend that is all public holidays but no weekend and iv) no holiday that is no holiday at all. Table 2 presents the minimum Npv and Nbat obtained for the charging stations located in offices parking with different number of holidays. It is clear from the results that fewer modules and batteries are required for a station with more number of holidays. This is obvious because more holidays provide a higher opportunity to sell entire PV energy during the absence of EVs in office parking. The last column of Table shows the value of fitness function (profit in USD) at which the PSO is converged. 6Results and discussion

Fig. 5. Operation during the idle scenario.

purchased from PV to feed EV, ESU and the grid. The factor (Gd2EV_ENR þ Gd2ESU_ENR) shows the amount of energy purchased from the grid at time t to charge the EV and the ESU. Similarly, the selling price is the sum of the energy sold to EV, grid and ESU. It is given as

Steng

The modeling and performance of the PV-grid system with the proposed scheme is simulated in MATLAB [77]. The objective is to test the resiliency of the algorithm in maintaining an uninterrupted charging at a constant price under various operating conditions. For illustration, the following resiliency tests are carried out for uninterrupted charging at a constant price: with varying weather conditions (normal and abnormal days) at parity and below parity levels (PL, 0.83PL, 0.33PL) without incurring economic losses by charging station when ESU_Pr or higher than PV_Pr

8 ðPV2EV ENR þ ESU2EV ENR þ Gd2EV ENRÞt ðPV Pr 1:0Þ > > < þðPV2Gd ENRÞt ðGE Prt 0:1Þ ¼ > þðPV2ESU ENRÞt PV Pr > : þðGd2ESU ENRÞt GE Prt

(26)

6.1. Resiliency during normal weather In (26) the factor (PV2EV_ENR þ ESU2EV_ENR þ Gd2EV_ENR) is the amount of energy sold to EV. This energy is sourced from PV, ESU and grid, respectively. The profit is simply the difference between selling and purchasing prices [32]. Since all the energy prices are taken in cents, the amount of profit is t Profit ðcentsÞ ¼ Steng Peng

(27)

By manipulating (25)e(27) and converting the cents to dollars, the total profit J is

Fig. 6 shows the charging profile of 150 EV on a specific winter day (data obtained for 15/01/2013) under the condition of parity, i.e. PV_Pr ¼ Avg_GE_Pr. The rise and fall of the PV_Pwr in the plot (a) is quite smoothdwhich indicates that the irradiance is not interrupted by a sudden change in weather such as rain, snow, large passing cloud etc. This shows that the subject day is with normal weather conditions. During the morning (before hour 10) and evening (after hour 18), the irradiance is low and the PV_Pwr is insufficient to charge all the EVs by means of PV2EV mode. In this situation, the EV_Dmd is ful-

93 28 ðPV2EV ENR þ ESU2EV ENR þ Gd2EV ENRÞt ðPV Pr 1:0Þ > > > > < =7 6 þðPV2Gd ENRÞt ðGE Prt 0:1Þ 7 6 7 6 þðPV2ESU ENRÞt PV Pr > T 6> > > 7 1 X : ; 7 6 þðGd2ESU ENRÞt GE Prt J¼ 7 9 8 100 t¼1 6 7 6 < ðPV2EV ENR þ PV2ESU ENR þ PV2Gd ENRÞ PV Pr = t 7 6 5 4 þðESU2EV ENRÞ ESU Pr t ; : þðGd2EV ENR þ Gd2ESU ENRÞt GE Prt The energy factors of PV and ESU in Eq. (28) include the number of modules in array and batteries in the storage unit. This is clear from Eqs. (8), (10) and (11). The objective function J in (28) is minimized to determine Npv and Nbat with the help of PSO like [66].

(28)

filled by activating additional modes, namely ESU2EV, and Gd2EV. The prioritization of the energy sources when GE_Pr ESU_Pr is in the following order: PV2EV, ESU2EV, and Gd2EV. The utilization of PV_Pwr is given first priority [27]. However, the battery is used if

392


Uninterrupted charging using PV, ESU and grid

Uninterrupted charging using PV and ESU

Uninterrupted charging using PV only

Valley-filling by ESU

Charging at constant price lower than Avg_GE_Pr

Fig. 6. Charging 150 EVs at PL on 15th January.

PV_Pwr is insufficient to fulfill the EV_Dmd. If the combined contribution of both sources is still inadequate, the grid power has to be utilized to maintain the continuous charging. This sequence can be clearly observed in the plot (b) during hour 9 to 10dwith the label “Uninterrupted charging using PV, ESU, and grid”. The transition from the ESU2EV to Gd2EV is determined by the present SOC of ESU. Once the ESU reaches the SOCL, the charging is taken over by the grid to ensure uninterrupted charging. Furthermore, the ESU is charged by surplus PV_Pwr or grid energy during off-peak conditions. For example, in the plot (c) during hour 1 to 2, the ESU is charged by the grid using valley-filling operation (because of GE_Pr < ESU_Pr). This is done by operating in mode Gd2ESU [28,78]. This is labeled as “Valley-filling by ESU”. Moreover, during hours 10 to 12, the ESU is charged again by the surplus PV_Pwr via the PV2ESU mode, as shown in plot (c). However, once the ESU reaches the limit of SOCU, the extra PV_Pwr is sold to the grid; this can be seen during period 11 to 18 in the plot (c). During hours 10 to 18, the PV_Pwr is more than EV_Dmd; this condition is termed as underload scenario. Here, the entire EV_Dmd is fulfilled solely by the PV energy via mode PV2EV. In plot (b), it is labeled as the “Uninterrupted charging using PV only”. However, during the evening, i.e. from hours 18 to 19, the PV_Pwr is less than EV_Dmd. This is defined as the overload scenario. Here, the EV_Dmd is fulfilled by two operating modes, namely the PV2EV and PV2ESU, as shown by the plot (b) marked as “Uninterrupted charging using PV and ESU”. The successful outcomes of the REMS can be seen by the resiliency of the charging price. This is evident from the plot (d) of Fig. 6, labeled as “Charging at a constant price lower than Avg_GE_Pr”. For the whole day, the Chrg_Pr remains constant as desired, i.e. 15.7 cents/kWh. This is irrespective of the type of energy source utilized and the variation in the grid electricity price. In addition, plot (b) indicates that the charging of EV takes place without interruption, despite the simultaneous utilization of all three power sources. Furthermore, the scheme results in the lowest possible burden on the grid. This is because, during peak conditions, the grid power is utilized as the last option. Moreover, plot (c) shows that there is no wastage of generated PV_Pwr, i.e. zero power curtailment. This is due to the fact that the

surplus PV_Pwr is consumed to charge the ESU and to feed the grid using modes PV2ESU and PV2Gd, respectively. It is worth mentioning that during the whole time of system operation, V2G and V2V operations are not involved at all. 6.2. Resiliency in abnormal weather Fig. 7 shows the operation for a specific day (19/02/2013) with abnormal weather conditions i.e. the intermittent solar irradiance. This intermittency is reflected by the fluctuating PV_Pwr throughout the whole day, as depicted by the plot (a). However, it is evident from the plot (b) that the EV_Dmd is fulfilled without any interruption, despite these fluctuations. Moreover, plot (d) shows that charging is done with constant price irrespective of continuous variations in GE_Pr. Plot (c) shows that during hours 14 to 15, the valley-filling by ESU (Gd2ESU) takes place. It is circled as “Valley-filling by ESU”. It shows that valley-filling can take place whenever the conditions are met, irrespective of time of the day. It is worth noting that, although the ESU has reached the SOCU limit during hours 17 to 18, the EVs are charged by the grid energy, instead of ESU because of GE_Pr < ESU_Pr. This is shown in plot (b) with label “Valley-filling by EV”. The valley-filling operation capitalizes on the low GE_Pr, while at the same time preserve the ESU energy for future use, particularly during grid peak conditions. Additionally, the minimal usage of ESU during off-peak conditions increases its battery life expectancy. This implies that during off-peak, the utilization of grid energy for charging EV is prioritized over the ESU. The results in the plot (b) and (c) of Fig. 7 show that both EV and ESU are benefitting from the low GE_Pr during the grid off-peak hours. It is important to note that operating valley-filling by EV while maintaining an interruption-less charging is a special feature of proposed REMS. This is because, in other research works, charging operation is interrupted and charging time is shifted to get the benefit of low price charging [28]. 6.3. Resiliency under different parity levels (PL) The previous results are obtained when charging station is


393

Valley-filling by EV


Fig. 7. Charging in abnormal weather on19th February.

operating at the parity level (for these simulations, the GE_Pr are obtained by multiplying normal electricity prices with PFM). However, it can be proved that the REMS is resilient in providing the constant price charging, even at low levels of parity. Fig. 8 presents the behavior of charging station for a specific day (15/ 01/2013) at below parity, i.e. PV_Pr is 0.83 times of Avg_GE_Pr (the GE_Pr are obtained by multiplying the similar normal electricity prices with 1.2 PFM). Below parity implies that the average grid electricity price is higher than the one generated by PV. Since the results in Fig. 8 are computed for the same winter day (15/01/2013), the charging profile is the same as in Fig. 6. However, for the below parity, there is a small shift in the valley-filling period, as shown in plot (c). It is labeled as “Delayed valley-filling by one hour”. For the parity case, the valley-filling takes place during the

hour 1 to 2, but for the case of lower parity, it is delayed by one hour, i.e. during hour 2 to 3. This is due to the fact that when operating at higher GE_Pr, the commencement time of the condition GE_Pr < ESU_Pr may change. It implies that with a higher GE_Pr, the valley-filling operation can be delayed or even suspended. Fig. 9 shows the results when PV_Pr is 0.33 times of Avg_GE_Pr, which means that the grid price is three times more expensive than the PV electricity. It is clear from the plot (c), there is no valley-filling operation at all, which means it is suspended completely due to much higher GE_Pr. It seems that by suspending valley-filling, the charging station (with a specific number of Npv and Nbat) is expected to face economic losses. But, at below parity, the higher GE_Pr compensates the suspension effect (reducing economic losses) of valley-

Delayed valley-filling by one hour

Fig. 8. Charging 150 EVs at 0.83 parity level on15th January.

394


hour 21 to 22, as shown in plot (c), with label “New valley-filling by ESU”. This happens because, with the new value of ESU_Pr (that is 18.4), the criteria of valley-filling (i.e. GE_Pr < ESU_Pr) is satisfied during hour 21 to 22. Under this situation, charging station can lose small profit because the energy stored (in ESU) at a higher price will be sold to EV at a lower price. This is necessary to keep the charging price constant at PV_Pr e 1.0 cents. However, this loss is insignificant because it is somehow limited by the additional valley-filling operations which are happening frequently due to higher ESU_Pr. Thus, even under this situation, the system is operated under the resilient control of REMS without facing economic losses which is proved in next sub-section. The results in Fig. 10 verify that with the variation of ESU_Pr, the resiliency of REMS is still sustained. This is clear from plots (b) and (d) that the EV charging is carried out without interruption with a constant price. Fig. 9. Charging 150 EVs at 0.33 parity level on15th January.

filling as the surplus PV energy is sold to the grid at a higher price. This is proven by the results shown in plot (d); the REMS is not affected by the variation of parity levels in providing uninterrupted charging at a constant price. This is another special dynamic feature of proposed REMS which shows that it is still applicable even the Avg_GE_Pr are three times higher than PV_Pr.

6.4. Resiliency with higher LCOE of ESU Previous results are computed when ESU_Pr is 1.7 cents less than PV_Pr. A simulation is carried out to check if the REMS is resilient with higher ESU_Pr than PV_Pr. Fig. 10 shows the charging profile of 150 EV at parity level for the same day (15/01/2013) when ESU_Pr is 18.4 cents/kWh. The results are similar to those presented in Fig. 6 but with the addition of another valley-filling operation during


6.5. Annual profit calculation Table 3 shows the annual profit of charging station for different number of office holidays, the number of EVs to be charged, parity level and ESU_Pr. These results are obtained by simulating the system (with sizes given in Table 2) for the period of one-year (2013). Since the system sizing is based on the period of lowest irradiance (01/09/2013 to 31/12/2013), the annual operation is guaranteed to be without loss. This is evident from the results given in Table 3. However, there is a small difference among the annual profits under different holiday conditions. This is due to the different ratio of Npv: Nbat. It should be noted that if Npv: Nbat is higher, the resultant profit is also higher as there is more PV_Pwr available to be sold to the grid. If the system is run at below parity, the annual profit is higher, compared to the operation at par parity level (PL). This is shown in the second last column of Table 3. This is due to the fact that at low grid parity, the Avg_GE_Pr is higher than the PV_Pr. Therefore, PV

New valley-filling by ESU

Fig. 10. REMS's results when LCOE of ESU is more than PV.


395

energy which is generated at a lower price (16.7 cents/kWh) is sold to the grid at higher prices (GE_Pr e 0.1), resulting in an increased profit. This implies that the REMS will be more profitable for the same system sizes in future (i.e. when the GE_Pr are higher than the PV_Pr). Similarly, the charging station is run with higher ESU_Pr. Annual profit under this condition is given in the last column. It is clear from the results that the charging station is still earning the profit due to the reasons already mentioned in earlier discussions. It is clear from above discussion that the REMS based operation of the PV-grid system remains lossless under different conditions of parity level as well as ESU_Pr. Additionally, this lossless operation is retained for the various system sizes obtained by PSO for a different number of holidays. 6.6. Comparison against standard grid charging The PV-grid based charging using REMS is compared to charging by the standard grid only [35]. For this purpose, the system is simulated with the standard grid as well as REMS based PV-grid systems. In both cases, the average per day charging prices for a same number of EV (150) with the same EV_Dmd under different cases of office holidays are calculated. It is clear from the results in Table 4 that the REMS is approximately 16.1e16.6% cheaper compared to charging using the grid only for different situations of office holidays. Table 5 shows the percentage decrease in average charging power load on the grid due to REMS compared to the standard grid charging. The results indicate that the load reduction varies from 93.7 to 98.9% for a different number of holidays. 7Addition of V2G technology in charging system In certain hours of the daydparticularly when the grid is extremely loaded, the tariff is very high. There is an opportunity for the EV that is standing idly in the parking lot to sell the energy from its battery to the grid [35]. The concept of selling the energy is known as the vehicle-to-grid (V2G) operation [62,79,80]. Although V2G process reduces the autonomy of interruption-less charging, it increases the monetary gains for both the charging station owner and the EV users; thus duration for the return on investment (ROI) can be reduced. The transfer of power from EV to the grid is carried out using the bidirectional dc-dc charger and the inverter, as highlighted in Fig. 11.

Fig. 11. V2G operation of PV-grid EV charging system.

To incorporate the V2G, certain modifications need to be made to the REMS algorithm. A new mode of operation (Mode 7) is introduced to quantify the amount of energy transferred from EV to grid. It should be equal to the energy available from the EV, i.e. the energy that is obtained by discharging the battery from SOCmax (80%) until SOCmin (10%) [62]. The energy is purchased from EV battery at a higher price (PV_Pr þ 2 cents/kWh) compared to the fixed charging price imposed by the station (PV_Pr e 1 cents/kWh). This incentive is recommended to attract the participation in V2G operation, as suggested by several researchers [61,62]. Since V2G is operated during grid peak hours, the valley-filling cannot be executed at the same time. This is because the valley-filling takes place during off-peak hours. It is important to note, the inclusion of V2G operation does not mean that the ESU can be eliminated from the system. The latter is still needed to store surplus PV energy (in the absence of EVs) for future use, i.e. to charge vehicles during low solar irradiance or grid peak hours. However, the size of the battery banks in the ESU can be greatly reduced. 7.1. Operation of charging station with V2G The objective of the simulation is to investigate the improvements in the payback schedule (and ROI) when V2G is incorporated into existing PV-grid charging system. As discussed previously, the overall operation is divided into four main scenarios. In the case of V2G, the idle and no load scenarios remain same due to the absence of EV to be charged. On the other hand, the overload

Table 4 Charging prices by REMS and standard grid. Office holidays

Per day average charging price by (USD)

Reduction in charging price using REMS (%)

Charging using standard grid only

REMS charging

All Holidays One Weekend No Weekend No Holiday

77.8 78.2 78.1 77.7

65.1 65.2 65.2 65.2

16.3 16.6 16.5 16.1

Table 5 Average grid burden by REMS and standard grid. Office holidays

All Holidays One Weekend No Weekend No Holiday

Average EV charging load on grid during peak hours (kW) Standard grid charging

REMS charging

15.6 19.0 21.7 22.4

1.0 1.0 0.4 0.2

Reduction in EV load on grid using REMS (%)

93.7 94.9 98.2 98.9

396


Overload Scenario EV_Dmd > PV_Pwr

Chrg_Pr = PV_Pr 1.0 Mode 1 PV2EV @ Chrg_Pr EV_Dmd NotFull valley-filling

V2G Operation

No

GE_Pr < Chrg_Pr

Mode 7 V2G @ GE_Pr– 0.1 EV_Dmd NotFull No

SOC > SOCL

Yes

Mode 3 Gd2EV @ Chrg_Pr EV_Dmd Full Yes No

No

(EV_Dmd PV_Pwr) > Avl_ESU_Pwr

Mode 2 ESU2EV @ Chrg_Pr EV_Dmd Full SOC < SOCU

Yes

Mode 2 ESU2EV @ Chrg_pr EV_Dmd NotFull SOC = SOCL

SOC < SOCU

Yes


Profit Calculation

Fig. 12. V2G based operation during the overload scenario.

(EV_Dmd > PV_Pwr) and underload (PV_Pwr EV_Dmd) are different because the EVs are assumed to be participating in the V2G operation. The modified flowcharts of the overload and underload scenarios are shown in Figs. 12 and 13, respectively. Note that the V2G (Mode 7) is activated only during grid peak hours. This implies that the valley-filling and V2G operations cannot be carried out simultaneously because the latter is initiated only during the off-peak hours. Since the V2G operates when the tariff is very high, the profit (obtained by the charging station) increases automatically. The extra revenue allows for the duration of the ROI to be reduced. Furthermore, Fig. 12 suggests that after the V2G is executed, the EV_Dmd may or may not be fulfilled. If Avl_ESU_Pwr > EV_Dmd, the EV battery will be fully charged. On the other hand, if the ESU energy is not sufficient during low PV_Pwr and grid peak hours, the EV_Dmd will not be fully satisfied. It means, by incorporating V2G, the interruptions in charging process may or may not be observed. For the underload scenario (as shown in Fig. 13), the EV_Dmd is fulfilled directly by the PV_Pwr. Thereafter, if the V2G is executed when the ESU is at low SOC and the PV_Pwr is insufficient, the EV demand cannot be fulfilled. Therefore, it is clear from the above explanations that under certain conditions, the EV_Dmd is not satisfied. Consequently, there will be undesirable interruptions in the charging process. As a result, the autonomy of interruption-less charging is no longer guaranteeddwhich can be considered as one

of the trade-offs of V2G operation. Despite this drawback, V2G remains as an attractive option for the participants who can get benefit from the incentives given by the station owner. 7.2. Results of V2G operation The results for the system that operates without V2G for 15 Jan and 19 Feb 2013 are (previously) shown in Figs. 6 and 7. The simulations that incorporate V2G for the same dates are presented in Figs. 14 and 15, respectively. As can be observed in the plot (c) of Fig. 14, the power is transferred from EV to the grid when the EV_Dmd is low. This is because, at this time, the vehicle can afford to sell its stored energy to the grid. Moreover, the plot (d) shows that during office hours (on a specific day), the grid remains overloaded. Therefore, no valley-filling by EV (Gd2EV) is realized. However, valley-filling by the ESU (Gd2ESU) takes place at night, i.e. during hour 1 to 2 as shown in plot (c). Furthermore, plot (d) indicates that despite of interruptions, the charging price remains fixed at the set value during the office hours. In Fig. 15, plot (b), the valley-filling by EV is carried out during hours 17 and 18. This is because the grid is at off-peak conditions, as shown in plot (d). However, during the hour 18 to 19, the grid is again overloaded. Since the vehicles are present in the office parking during this time, the stored energy in EV battery is transferred back to the grid via


397

Underload Scenario EV_Dmd > 0 AND EV_Dmd