The design of a solar-powered transportation system requires topographic, transportation as well as solar radiation data. First, the energy demands for the routes are identified using an average figure for energy consumption per distance unit, i.e. 1.35 kWh km⁻¹ (Jang, 2018; Gao et al., 2017). Although this is a crude estimate, it is reasonable, since, inside the campus, there is no significant stop and go traffic neither large road segments with steep slopes. Route operational characteristics are collected through the website of the Athens Public Transport Operator (http://www.oasa.gr/, last access: 21 November 2019). The route is circular and connects the metro line to the university comprising 19 bus stops, 10 of which outside the campus (Fig. 1). Three buses run 239 d a year and the number of bus routes operated daily is 66 covering a total of 495 km. Therefore, in terms of route energy requirements, the total amount is approximated as 495 km d${}^{-\mathrm{1}}\times \mathrm{1.35}$ kWh km⁻¹ = 668.25 kWh d⁻¹. Additionally, elevation data (Digital Elevation Model – DEM) are acquired from Hellenic Cadastre at a spatial resolution of 5 m (Fig. 2). Last, available solar panel and electric bus technologies are examined. The chosen configurations for the electric buses in terms of battery and charger types are based on the data reported from recent literature (Lajunen, 2018; Liu and Song, 2017; Laizāns et al., 2016) and are analyzed in Sect. 4.

For the siting of the PV panels at the NTUA campus, solar radiation is calculated at the spatial scale of the DEM using the solar radiation analysis tool of ArcGIS, taking into account topographic effects (Mamassis et al., 2012). The Area Solar Radiation tool of ArcGIS is based on methods of the hemispherical viewshed algorithm developed by Fu (2000) and Fu and Rich (2002). The method consists of summing direct and diffuse radiation for every location on the topographic surface, using a centroid at zenith angle and azimuth angle, at the case of direct, and corrected by the gap fraction and angle of incidence, at the case of diffuse radiation. The DEM of the NTUA campus used in the analysis is expanded to include adjacent hills that cause shading to the campus area at the morning hours. For the preliminary stage of the feasibility analysis, the clearness index is assumed constant disregarding its temporal variation (i.e. no diurnal, seasonal or annual fluctuations) and with uniform spatial distribution over the area of the campus. We consider the clearness index equal to 0.6 as estimated by the expected hourly value of the clearness index during daylight over the campus (Koudouris et al., 2017).

Figure 2(a) Digital elevation model (DEM) of the campus. (b) The total solar radiation at the campus area. (Base Map Layer: Bing Map v.7, Bing Satellite Base Map).

The surface solar energy (kWh) is calculated as:

where S is the solar irradiance during the days of operation (kWh m⁻²), i.e. 239 d yr⁻¹, and A (m²) is the direct area in which solar panels are installed in each case. Photovoltaic energy generation is calculated as:

where the capacity factor (CF) of the conventional solar panels, which ranges between 13 %–17 % (WEC, 2016) and for our study is assumed as 14 % (following Jewel et al., 2006), while that of solar roads as 7 % due to decreased performance from the use of thick protective glass, shading from moving cars and the accumulation of dust, as explained in Sect. 4.3. The resulting raster layer of solar radiation at the NTUA campus (Fig. 2) is used to select the locations for installing solar panels in each scenario, where possible, among the available areas (Table 1). Out of the three scenarios examined, scenario 3, i.e. solar roads, albeit its lowest efficiency in terms of harnessing solar radiation, allows the greater flexibility in the selection of locations for the installation of solar roads, due to the large availability of road surface at the campus. In particular, the top 2400 m² in terms of solar radiation are selected out of the total road area of NTUA, which is 37 970 m² (Fig. 4). In the other two scenarios, siting of solar panels is restricted by existing locations of bus stops, in scenario 1, and by the availability of large and unused open spaces, in scenario 2. In the case of solar bus stops, a slightly higher average solar radiation is achieved by selecting the 5 out of 9 existing campus bus stops that receive the maximum radiation.

Table 1Total area used, direct PV area (A), total solar irradiance (S), total surface solar radiation (S_sur) and total solar energy production (E) during days of operation per year for the three scenarios.

Download Print Version | Download XLSX

Three scenarios are explored for the collection and use of solar energy: (1) solar panels installed on the top of bus station shelters (2) solar panels installed at unused open spaces and (3) road segments constructed by photovoltaic material, commonly referred to as solar roads. In all three cases, it is assumed that a stationary charger at the terminal is required for the vehicles to start the routes fully charged (Gao et al., 2017). In the following sections, we analyze of each scenario in terms of its technical configuration and energy produced.

4.1 Scenario 1: Solar bus stops

In this scenario, we explore the emerging concept of solar bus stops, i.e. bus stops collecting solar energy either from their shelter constructed from PV material or from PV panels installed in close proximity to the bus stop and transporting the energy to buses via induction (Franken and Meijer, 2014; Kawashima and Fujioka, 2008). This charging configuration is referred to as opportunity charging, as buses charge during dwell times at stops through high-power chargers (Iliopoulou et al., 2019; Lajunen, 2018). In this case, we assume that candidate stop locations must be within the campus, while the stops exposed to the highest amount of solar radiation must be exploited. For this purpose, based on the resulting solar radiation map (Fig. 1), we select the stops located at the areas exposed to the highest amount of solar radiation. These are five stops namely those at Gate Kokkkinopoulou, Gate Katechaki, School of Civil Engineering, School of Naval Engineering and Student Housing. The selected stops are also associated with high passenger loads (Asimomiti et al., 2019), thus they constitute advantageous locations for charging during dwell times. The short route length and thus, the corresponding low energy requirements, allow for the deployment of low-capacity high-power batteries of 30 kWh (Jungmeier, 2017). Further, we assume high-power charging of 200 kW and 60 s recharging during dwell times, which are typical values for the concept of opportunity charging (Jungmeier, 2017; Jang, 2018). Under these specifications, a typical energy consumption and recharging scheme for two consecutive routes is shown in Fig. 3, assuming 97 % charger efficiency (Gao et al., 2017). Vehicles need to recharge approximately every hour, as the average cycle time is 40 min with small daily fluctuations while as shown in Fig. 3 charging is required every other route. The solar roof of each bus stop is assumed to be 30 m² (10 m × 3 m), a value within the literature range (Kawashima and Fujioka, 2008). Based on the average annual solar radiation at the campus (1109 kWh m⁻² yr⁻¹), the solar energy per year at the area of all bus stops is 166 MWh. Considering the solar panel efficiency rate (14 %), the production of electricity during the year amounts to 16 MWh in total, or approximately 16/239 = 0.067 MWh d⁻¹ on average for each day of operation. These calculations are summarized in Table 1. Obviously, the solar panels at the bus stops do not provide enough energy to autonomously operate the buses and a total of 572 kWh d⁻¹ of supplementary electricity from the grid would be required on average.

Figure 3(a) Energy consumption and recharging route profile. (b) The five selected bus stops receiving the highest solar radiation amount (Base Map Layer: Bing Map v.7, Bing Satellite Base Map).

4.3 Scenario 3: Solar roads

The third scenario considers the installation of solar panels along segments of the NTUA campus road network, in order to collect solar radiation and dynamically power the buses as they move along them using induction (Karataraki et al., 2019; Venugopal et al., 2018). To select these segments, total annual solar radiation on the campus roads is mapped using GIS software according to the process described in Sect. 3. Figure 4 illustrates the annual radiation for the PV area along the route.

Figure 4(a) Solar radiation at the open space PV area (scenario 2) and (b) at the solar roads surface (scenario 3). (Base Map Layer: Bing Map v.7, Bing Satellite Base Map).

The total road surface of the NTUA campus, shown in Fig. 4, is 37 970 m². Based on the results shown in Fig. 4 on road solar radiation, road segments of approximately 300 m in length and 8 m in width (the total area is 2400 m²) are selected, using two criteria: high incoming solar radiation and low shading from buildings and trees. We also note that the selected areas have limited traffic and are not used for short-term parking. Not considering effects from dust accumulation and dynamic or static shading, the overall annual efficiency of similar set-ups has been estimated as 8.6 % (Shekhar et al., 2015). In this work, in order to account for shading from moving cars and increased concentration of dust, the capacity factor is slightly reduced to 7 %. However, it has to be noted that the capacity factor of such configurations has been reported as low as 4 % (https://www.independent.co.uk/environment/solar-panel-roads-energy-bad-idea-renewable-climate, last access: 21 November 2019). The total number of panels installed in this area is thus 1238 (2400/1.937), while the solar energy per year at the area is 2069 MWh. So, in one year 145 MWh can be produced (Table 1). For 239 d of operation, the solar roads could on average generate 145/239 = 0.607 MWh d⁻¹.

4.4 Preliminary capital cost analysis and scenario comparison

Indicative capital costs for each of the three scenarios are presented in Table 2, using the PV cost figures found in IRENA (2012) and assuming the costs for the electric buses and charging devices according to the analysis of Lajunen (2018). Namely, for the scenario of solar bus stops, costs for electric buses and dynamic charging devices are taken as EUR 350 000 and 250 000, respectively. Five chargers are required, plus one at the terminal. Regarding battery costs, the value of EUR 100 kWh⁻¹ (Liu and Song, 2017) is assumed for low-capacity batteries used for the vehicles (3 batteries of 30 kWh) and for storing solar energy at the specific bus stops (5 batteries of 10 kWh each). For the scenario of PV panels in open space, one overnight charging device is required, with a unit cost of EUR 20 000, and high-capacity batteries for the buses (180 kWh) and for storing the PV area energy (700 kWh) at a cost of EUR 600 kWh⁻¹ (Laizāns et al., 2016). Finally, for the solar roads' scenario, we assume four fast charging devices for the road segments plus one at the terminal, and the same battery configuration as in the solar bus stop scenario. Concerning the cost of PV panels per m² (including installation and panels), a value of EUR 4465 kW⁻¹ is assumed for conventional PV panels (IRENA, 2012) in open space, and twice as much (EUR 8930 kW⁻¹) for PV panels in bus shelters. Finally, a value of EUR 1785 m⁻² or EUR 11 905 kW⁻¹ of installed capacity is used for the cost of solar roads, based on the costs reported for the recent large-scale solar road application in France (https://theconversation.com/solar-panels-replaced-tarmac-on-a-road-here-are-the-results, last access: 21 November 2019). Total costs are summarized in Table 2. At this stage lifecycle costs are not taken into account although it is expected that the solar roads' technology has the highest ones due to maintenance requirements, since the panels on the road can be easily broken or ruined (Karataraki et al., 2019).

Overall, the open space scenario is the most advantageous, in terms of capital cost per unit energy generation, with the primary drawback being the sub-optimal exploitation of space since the area of installation cannot support other operational uses as in the other two cases. Evidently, the solar bus stop technology is overly capital intensive, at the scale examined, as the cost for batteries, chargers and vehicles is prohibitive for a system with such low energy generation. Furthermore, there are not enough suitable stops in order to integrate this technology at a larger scale. Solar road technology on the other hand, suffers from high infrastructure cost and lower capacity factors, the combination of which leads to high capital costs per energy unit produced, even though the scale of the layout examined is comparable to the open space scenario. The total capital costs are EUR 171 378 per MWh of energy generated for solar bus stops, and this drops to EUR 10 891 for PV panels (scenario 2, which is the most advantageous). It should be also noted that this comparison refers to solar energy production and does not consider the power transfer efficiency of each charging scheme. In this case, greater discrepancies would arise between static and dynamic charging, as in the former case efficiencies up to 97 % can be attained (Gao et al., 2017) versus 80 for the latter (Jang, 2018).

Table 2Capital cost analysis.

Download Print Version | Download XLSX

With both preliminary capital cost investigation and energy generation estimation converging to the prevalence of scenario 2, additional investigation is carried out for PV placement in open space, in terms of the relation of the scale of the system, i.e. the size of the PV surface, with profit and reliability. For this, a model for the daily operation of the system is developed based on the layout described in the corresponding section. Namely, this consists of an area of 3000 m², in which 375 kW of PV panels are installed, with a 700 kWh battery for the charging station, used for overnight charging of the buses. The total daily consumption of the electric buses is 668.25 kWh and the system is connected to the grid, buying for EUR 0.18 kWh⁻¹ when in deficit and selling for EUR 0.08 kWh⁻¹ when in surplus¹. To simulate the operation of the system on the daily scale, we use historical data of daily solar radiation from the meteorological station of NTUA spanning from 17 Nov 1998 to 5 Sep 2009 (https://openmeteo.org/stations/1334/, last access: 21 November 2019). From the simulation, we identify the occurrences where the energy demand was not met and thus energy is bought from the grid, as well as the occurrences in which there was energy surplus and therefore energy was sold to the grid. Accordingly, we estimate the buy-sell ratio, and the estimated yearly profit from the operation, both shown in Fig. 5. It is evident that the system becomes profitable only when the installed PV surface exceeds 206 m² while it becomes nearly autonomous, i.e. achieving a 1 % buy-sell ratio, at 600 m². Clearly, for a given configuration, the degree of autonomy of the system is mainly dependent on the natural variability of solar radiation, whereas its profitability is additionally highly dependent on the variable cost of buying and selling to the grid, which here is simplistically assumed to remain constant over the years. Therefore, in order to reach autonomy, a much smaller system could be employed, e.g. 600 m², yet if profitability is a target, both the life-cycle cost of the system and the variability of the energy market should be considered.

Figure 5(a) PV surface vs expected yearly profit and (b) PV surface vs. buy-sell ratio.

Download

Evidently, there is a lot of potential for the implementation of such systems in places receiving high solar radiation. Yet, more research is required in order to better account for the variability induced by the stochastic nature of solar radiation. Here we have used observed historical solar radiation data from a station inside the campus spanning approximately 10 years of daily data. We note that in general, the spatio-temporal variability of the hourly clearness index may be significant since it depends on local topographic characteristics (mostly affecting scenarios 1 and 3) and on the temporal variability, although we do not expect the spatial variability to be significant at such a small scale. The temporal variability is triggered by the short-term high autocorrelation as well as by the long-term behaviour identified from the time series analysis (Dimitriadis and Koutsoyiannis, 2015) of solar irradiation both over the area of interest and globally (Koudouris et al., 2017, 2018). A proper measure to express the strength of the long-term persistence of a stochastic process is the Hurst parameter, (Hurst, 1951; Koutsoyiannis, 2010). Interestingly, the Hurst parameter for the clearness index over the area of interest is estimated around 0.8 (Koudouris et al., 2017) indicating strong long-term persistence of the process and thus, a relevant model will be developed in future research to assess the degree of uncertainty induced from the variability of the clearness index based on synthetic solar radiation timeseries longer than the historical data used here.

However, even the single simulation of the system based on historical data is reflective of the effect of the scale of the system on both reliability, pertaining to minimizing the energy required from the grid, and profit, estimated from the process of buying and selling energy to the grid and assuming the present energy prices. Additional considerations on the scale of the system should take into account both potential conflicting uses of the available spaces as well as aesthetic issues relating to the potential degradation of the landscape (Sargentis et al., 2019).

Further research is required in order to quantify the whole life-cycle cost of the project, taking into account specifications about the batteries life-cycle as well as maintenance costs, which are expected to be significant particularly for solar road technology. Level of transit service provided by the different systems could also be investigated further (Papantoniou et al., 2019). For example, in scenario 1, additional stationary time in bus stops is required for charging in contrast to scenario 2, in which overnight charging is utilized. Efficiency of service was maintained in scenario 1 through the selection of stations with high on-off load for charging, but as part of a large investment on the campus transport system we believe that improving the efficiency of the system should also be investigated as well as the trade-off between heavier batteries for overnight charging and bus capacity (Vlahogianni et al., 2018).

Last but not least, it should be noted that the results are highly dependent on the selected capacity factor, which is shown to often deviate from theoretical values, especially in renewable energy technologies (Boccard, 2009; Ioannidis and Koutsoyiannis, 2019).

Historical solar radiation data for the NTUA campus can be accessed at https://openmeteo.org/api/stations/1334/timeseries/231/data/?fmt=csv (last access: 21 November 2019; National Technical University of Athens, 2019). Elevation data (Digital Elevation Model – DEM) can be accessed upon request from the Hellenic Cadastre at http://www.ktimatologio.gr/sites/en/aboutus/Pages/qoQhyNCvtozm6ajS_EN.aspx (last access: 21 November 2019; Hellenic Cadastre and Mapping, 2019). The data are only provided for free to the Greek Public Sector.

Study conception: TI, PD, GK, RI; data analysis: RI, TI, CI, LK; data collection: LK, AP, MEM, MEA, NP; manuscript preparation: RI, TI, CI, LK, NM, PD; study supervision: DK, NM, KK, EIV, CP. All authors read and edited the manuscript.

The authors declare that they have no conflict of interest.

This article is part of the special issue “European Geosciences Union General Assembly 2019, EGU Division Energy, Resources & Environment (ERE)”. It is a result of the EGU General Assembly 2019, Vienna, Austria, 7–12 April 2019.

This research has been supported by the Eugenides Foundation (Scholarship for doctoral studies in NTUA grant).

This paper was edited by Sonja Martens and reviewed by Dylan Ryan and one anonymous referee.