Three cases of study are considered; the first as a baseline, and the following represent two possible future scenarios with solar and H₂ integration.

• Case 1: EV charging directly from the electric grid. This case considers the incorporation of an ultra-fast charger connected directly to the electricity grid.

• Case 2: EV charging using a H₂ fuel cell. In this scenario, the ultra-fast charger obtains its energy from a fuel cell, assuming an ensured supply of H₂.

• Case 3: EV charging using a fuel cell with H₂ produced with solar PV. This case is similar to Case 2, with the distinction that both the housing complex and the residential building generate the H₂ required to power the charger using solar energy. If the H₂ produced is insufficient to meet the charging demand, the electricity from the grid will be used as a supplementary source.

To carry out a more representative study, three types of EVs were selected for the current analysis: a city car, an SUV, and a sedan. According to the Zero and Low Emission Vehicle Sales Report in year 2024 in Chile [26], the Tesla Model 3 was the best-selling EV up to June 2024. The Model 3 is the rear-wheel version, with a battery capacity of 57.5 kWh. In the same report, the Kia Niro ranked eighth in sales. Although it was not the best-selling SUV, this model was selected due to technical specifications availability. Finally, the Fiat 500e was chosen as the city car.

After selecting the vehicles to be analysed, their behaviour under ultra-fast charging conditions is simulated. To accomplish this, the behaviour of a single cell is simulated under different charging strategies (CCCV: Constant Current Constant Voltage, or MSCC: Multiple Stage Constant Current), starting from an initial state of charge (SOC) of 10% and reaching a final SOC of 80%. The basis to perform each charging simulation assumes that each battery pack has a specific cell configuration to deliver the nominal capacity and, and each cell is simulated using the PyBamm library [27], employing the Doyle–Fuller–Newman (DFN) model to represent the electrochemical behaviour of lithium-ion cells [28]. Following this logic, four configurations of CCCV and MSCC ultra-fast charging strategies are applied, and their results are then averaged to obtain a representative ultra-fast charging profile for each EV.

To select the characteristics of the residential complex to study, qualitative features of EV adoption are considered. Since EV prices are still prohibitive for a part of the population, it is more likely that their adoption will initially increase in areas with higher income levels. In Santiago, Chile, the five communes with the highest income per capita correspond to Vitacura, Las Condes, Alhué, Providencia, and Lo Barnechea. Since this study focuses on houses and apartment complexes, it was necessary to select a location that contains both types of residential arrangements. Among these five communes, Las Condes offers the best proportion of condominiums and apartment buildings, making it the most suitable choice for the study.

The next step involved conducting a survey of the area to select a representative apartment building and condominium, including the number of flats, houses, and parking spaces. A field visit was carried out in Las Condes to collect data, resulting in a sample of four condominiums and four apartment buildings. In addition, Google Earth was used to estimate the roof surface areas of the properties, which is essential to determine the available space for installing solar panels in Case 3. The final selection consists of a condominium with 10 houses, with a total of 1.539,61 m² of roof space, and a residential apartment building comprising 60 flats and 120 parking spaces, with a total roof space of 1.185,92 m².

From this selection, the charging infrastructure was defined in an ad-hoc manner for the scenarios under analysis. In the housing complex, a single charging station will be considered with a scheduled usage, allowing the charging of 10 EVs (one per household). In the residential building, 10% of the parking spaces will be equipped with charging points, equivalent to 12 stations. These will also operate under scheduled usage to ensure that no vehicle remains fully discharged. Furthermore, each flat in the building is assumed to own one EV, resulting in a total of 60 vehicles requiring charging. Currently, there is no local public data on this type of infrastructure; therefore, a more realistic and optimal determination of the number of chargers is left for future work.

Once the residential building characteristics have been defined and the vehicle charging profiles have been simulated, a monthly and subsequently annual charging demand plan is generated. To establish a schedule for the use of the charging infrastructure, three annual driving-distance scenarios for each vehicle are considered based on local statistics [29], as shown in Table 2. The corresponding daily driving distance is then calculated, along with the resulting change in SOC. This makes it possible to determine how many days of use can occur before recharging is required, as well as the amount of energy needed to restore the battery to 80% SOC. Based on these factors, the annual distance travelled and the charging infrastructure available in each residential building, an annual charging plan is developed.

A one-month charging community schedule was then created, including the number of vehicles and type, charging frequency, daily number of vehicles connected, and the duration of charging shifts for both the housing complex and the residential building. The scope was subsequently expanded to a one-year horizon. For Cases 2 and 3, the equivalent amount of hydrogen required to meet this demand was calculated using a cell conversion factor.

For Case 3, the hydrogen production using solar energy was modelled, based on a simulation on pvlib [27]. The code utilised typical meteorological data, calculated solar position based on the geographic coordinates, given a standard 30° fixed-tilt and north-facing configuration, and determined the PV output power considering technical parameters. PV arrays were defined for both the housing complex and the residential building according to the available roof surface.

The PV energy was calculated monthly, then converted into H₂ production using the electrolyser’s conversion factor, and finally into usable charging energy through the charging cell conversion factor. The annual simulation informs the total yearly energy generation and daily averages, as well as additional information for Cases 2 and 3: annual H₂ production, daily H₂ output, and electricity converted from H₂. These results were compared against the previously calculated annual demand to assess whether H₂-based energy production would be sufficient to meet EV charging requirements.

A market survey was conducted to select the equipment required for each case study. The chosen equipment had to meet the established technical requirements, with priority given to small units due to space limitations in the infrastructures under analysis. Price quotations were also obtained, necessary for the subsequent economic evaluation.

In Case 1, where EVs are charged directly from the electricity grid, the only equipment required is an ultra-fast charger. Case 2 introduces H₂ as an energy storage and conversion medium, requiring a charger, a storage tank, and a fuel cell. In this configuration, the H₂ stored in the tank is converted into electricity by the fuel cell, which is then delivered to the EVs through the charger. Case 3 incorporates renewable generation, with the necessary equipment consisting of the PV system, an electrolyser, a compressor, a storage tank, a fuel cell, and a charger. Here, solar energy powers the electrolyser to produce H₂, which is compressed, stored, and later converted into electricity by the fuel cell before charging.

The equipment selection process considered the technical specifications required for implementation in both the housing complex and the residential building. Challenges arose due to H₂ being a relatively new technology, with most available equipment designed for industrial applications rather than residential projects. Furthermore, as this was a feasibility study, companies contacted showed little interest in providing detailed quotations, given the absence of real purchase intentions. Consequently, bibliographic references and secondary sources were used to estimate prices and identify equipment suitable for the specific needs of the project. Technical details regarding the specific equipment units selected are available in the Appendix.

The economic evaluation began with a detailed analysis of the investment costs required for project implementation. These costs were determined from the equipment selection stage, as the investment consists solely of acquiring new equipment. The next step involved establishing the price of electricity (kWh) from the grid and at existing EV charging apps, and an estimation of the annual percentage increase in energy prices. Fixed and variable costs were then considered, with only the fixed cost of annual equipment maintenance included. Note that an approximate conversion of 1 USD equal to 1000 CLP was assumed for simplification purposes.

Cash flow models were developed for each case study in both the housing complex and the residential building. Discount rates and corporate tax values were defined, and the need for reinvestment was assessed based on equipment lifetime and the cash flow period. Six cash flows were calculated and analysed. To better understand their behaviour, sensitivity analyses were performed on the most relevant variables, such as the selling charging price, with results used to evaluate their influence on the Net Present Value (NPV).

The same cash flows were then recalculated under a financing scenario, assuming 60% of the project cost was funded. Two loan repayment methods were compared ‒ constant instalments and constant amortisations ‒ to determine which was more suitable for the study. An additional case was added to consider the effect of including electricity price projections based on tariff data from recent years.

Finally, key economic indicators were calculated. The Levelised Cost of Energy (LCOE) was determined to establish the minimum price per kWh required for profitability, calculated only for Case 3, as it is the only scenario involving energy generation. In addition, the Levelised Cost of Hydrogen (LCOH) was calculated, representing the minimum price per kilogram of H₂ required for the project to be viable. This indicator was also applied exclusively to Case 3, where H₂ production takes place.

The EV charging profiles were generated in python using the pyBAMM library with each set of EV parameters, resulting in charging power curves that depend on the SOC. The solar PV power generation profiles were generated in python using the pvlib library with each configuration for the housing and apartment complexes. Finally, the full technical and economic analyses utilised these profiles to perform energy balances and cash flows in spreadsheets.

The modelled charging characteristics for each EV are presented in Table 3, which reports the time required to charge the battery from 10% to 80% SOC, the corresponding energy delivered, and the equivalent amount of H₂ needed to supply that energy. These indicators reflect the average characteristics obtained from the simulated charging models.

Figure 2 illustrates the characteristic charging schedule developed for the condominium of houses, which includes 10 EVs and a single charger. The charging requirement is determined solely by the energy consumed during the days of use, which is assumed as a constant behaviour along the year.

Figure 2.

Charging schedule for the housing complex

The analysis determined the charging frequency required to operate within the desired SOC range, preventing full battery depletion. Based on typical average annual travelled distances, an average daily battery consumption was estimated for each model. With a safety margin applied, all vehicles were found to require charging within a five-day cycle.

For the housing complex, which contains ten EVs, the distribution was set at three Fiat, three Kia, and four Tesla. To meet the five-day cycle, two vehicles must charge each day. A single charging station was scheduled to operate in two one-hour shifts at 6:30 pm and 7:30 pm, ensuring vehicles reached a final 80% SOC. These times were selected to align with typical residential working hours.

Naturally, users may deviate from the reserved timeslot for charging. The simulation has a certain margin for error since the charging timeslots are for 1 hour and the needed time to charge is around 35 minutes. If larger deviations were to occur, the daily charging demand may not be constant but have a stochastic behaviour that could be better described through its statistics. The worst case would correspond to a greater charging demand in a specific day, which could result from either charging for longer than the allocated time, arriving with a critically low SOC, or having more users one day than the reservations. These types of variations could result in a larger energy or power demand, which in turn increase the technical requirements of the system.

In the residential building, 60 EVs were evenly distributed among the three models (20 Fiat, 20 Kia, 20 Tesla). To comply with the five-day cycle, 12 vehicles were scheduled to charge daily. A single one-hour charging slot was defined at 7:00 pm, again chosen to suit residential contexts.

Throughout an entire year, subsequent cycles of charging were simulated, with initial SOC values that were estimated from the daily discharge rates and correspond to: 18% for Fiat, 32% for Kia, and 35% for Tesla. Vehicles were then charged from these levels up to 80%.

From these schedules, the daily and monthly energy demand was calculated, enabling estimation of the H₂ required to supply this energy. Extending the model to a full year, with the same charging schedules and EV distributions, allowed to calculate the total annual energy demand, being 21,432.2 kWh for the housing complex and 132,909.4 kWh for the apartments; equivalent to a H₂ provision of 1,286.1 kg and 7,975.4 kg, respectively, as shown in Table 4.

Case 3 analysed EV charging directly from a H2 cell powered by PV panels. Hydrogen production was modelled after the annual solar generation. A standard array with 205 W Kyocera modules [30] was considered in a surface limited to 50% of the available roof area in both residential settings. While the building achieved values close to this threshold, the housing complex used slightly less due to inverter capacity constraints, with the resulting PV systems shown in Table 5.

The energy generation and conversion model was then extended to a full year, incorporating seasonal variations. Annual results in Table 6 include total energy generation, daily averages, H2 production, and electricity derived from H₂, allowing a direct comparison with the EV charging demand, to evaluate feasibility.

Figure 3 and Figure 4 present the electric energy generation from H2 and solar sources, together with the charging demand. It can be observed that, in the case shown in Figure 3, which corresponds to the housing complex, the demand is met throughout the year, with an overgeneration exceeding 200 kWh during the summer months. In contrast, the opposite behaviour is observed for the apartment complex, as illustrated in Figure 4, where at no point during the year does the electricity generation from H2 and solar energy suffice to meet the EV charging demand.

Figure 3.

Monthly supply of H2 and EV charging demand for the housing complex

Figure 4.

Monthly supply of H2 and EV charging demand for the apartment complex

Not meeting the charging demand in the case of the apartment building stems from the assumed constraint of using only the 50% of available roof surface for a solar PV installation. While this constraint could be lifted, it would be unrealistic to achieve a 100%, which would still underserve the demand during winter months. This fact highlights the space limitation inherent in dense urban environments.

By calculating the energy generated by the PV system and the energy recovered through the conversion process, the overall process efficiency is obtained, which is approximately 35%. This indicates that when H₂ is used as an energy storage medium, around 65% of the initial energy is lost throughout the process due to the multiple conversion stages before the ultra-fast charger: Case 3 converts solar irradiance to power in PV cells, then that electricity is used to generate H₂ (with an efficiency of approximately 70%), which is stored in a tank, and then reconverted to electricity in a fuel cell (with an efficiency of approximately 50%), which is finally used for charging. In contrast, Case 2 only has one conversion stage: from supplied H₂ to electricity in a fuel cell, and finally, Case 1 does not perform any conversion since electricity is supplied by the electric grid.

This low efficiency is concerning, as it questions whether such technologies are indeed sustainable. However, evaluating sustainability is not simple. Given a future case in which the electric grid may become highly saturated or even experience frequent blackouts due to increased electrification, ultra-fast EV charging could be restricted, and H₂ conversion technologies could offer accessibility and resilience, both dimensions of sustainability. Eventually, multi-criteria evaluations could consider technical and accessibility trade-offs for a comprehensive sustainability assessment.

Table 7 presents the investment costs of the main equipment required to enable residential charging using H2. It is important to note that, among the three cases analysed in this study, the first case requires only an EV ultra-fast charger. In the second case, a H2 tank, a fuel cell, and the EV charger are required, whereas in the third case all the equipment listed in the table is necessary.

Table 7 shows that the costs associated with the apartments are significantly higher compared to the houses. In Case 3, the investment required for the building approximately doubles that of the housing complex. The main difference lies in the number and capacity of the ultra-fast chargers, as the apartment complex requires four 400 kW chargers, whereas the housing complex requires only two 200 kW chargers. The selection of the fuel cell, in this case, was based on the maximum charging power, which ensures effective ultra-fast charging. It is worth noting that if smaller equipment were selected, the feasibility of Case 3 could be improved; and since charging has varying power, there are times when the full capacity is not fully utilised. This could be a matter of future research, to evaluate configurations with lower investment costs, at lower levels of fast charging instead of ultra-fast charging.

Figure 5 and Figure 6 show the minimum energy charging sale price required for the project to achieve a positive Net Present Value (NPV). These cases assume an initial electricity price of 0.33 USD/kWh and a conservative 5% annual growth. The corresponding NPV at this sale price is also reported. Figure 5 presents the results for the three case studies in the housing complex, while Figure 6 shows the results for the apartment building. It is important to highlight that, in both cases, the highest NPV is obtained in Case 2, with sale prices similar to those of Case 1. In contrast, Case 3 requires a significantly higher sale price and results in a lower NPV.

Figure 5.

Resulting minimum energy charging price and NPV for each case studied in the housing complex

Figure 6.

Resulting minimum energy charging price and NPV for each case studied in the apartment complex

It can be observed that, for Case 1 in the housing complex, the required charging price is relatively close to the current reference values in commercial chargers and could therefore be considered feasible. However, in Cases 2 and 3, the required prices increase significantly, reaching approximately twice and up to five times the current market price, respectively. Given these high costs, H₂-based solutions for ultra-fast charging are unfeasible for the residential scale today, and their eventual deployment in the near future would require important support from subsidies.

For the apartment complex, the prices exhibit less variation across the three cases compared to the condominium. This behaviour is explained by the higher annual energy demand of the apartments, for which a marginal increase in the charging price has a more pronounced impact on the revenue. For Cases 1 and 2, the resulting prices remain close to the assumed market values, whereas for Case 3, the minimum required price rises to more than twice the current market price.

Figure 7 and Figure 8 present a sensitivity analysis conducted for Case 2, in which EVs are charged using a fuel cell and H₂ is stored in tanks. This analysis illustrates how the expected Net Present Value (NPV) of the project changes as four different parameters are varied: discount rate, annual grid electricity price increase, first-category tax, and loan amount. The behaviour of the NPV with respect to these parameters is consistent for both apartment buildings and condominiums. Specifically, the resulting NPV decreases as the discount rate and first-category tax increase, and the NPV increases as the annual grid electricity price increases and loan amount rises.

Figure 7.

Sensitivity analysis of the economic indicators for case 2 in the housing complex

Figure 8.

Sensitivity analysis of the economic indicators for case 2 in the apartment complex

From the sensitivity analysis shown in Figure 7 and Figure 8, it can be observed that although the discount rate influences the NPV, its impact is not significant. Regarding the first-category tax, its influence on the NPV is practically negligible. For instance, over a variation range of 10 percent, the NPV for the housing complex changes from −33.7 thousand to −34.1 thousand USD, and for the apartment complex from −74.5 thousand to −77.5 thousand USD. In contrast, the annual increase in the electricity charging price (per kWh) exhibits a strong influence on the NPV. If a 20% annual increase – similar to that observed in 2024 – is assumed, profits of up to 250 thousand USD for the condominium and approximately 1.4 million USD for the building could be achieved.

Finally, Table 8 presents the results for the two key indicators: the Levelised Cost of Electricity (LCOE) and the Levelised Cost of Hydrogen (LCOH) for both residential settings. These indicators reflect the economic feasibility of electricity and H₂ generation, respectively. They are calculated for Case 3, as it is the only case in which both H₂ and electricity are produced for charging purposes.

When analysing the LCOE, similar values are obtained for the housing complex and the apartment complex. This is because, although the housing complex incurs in higher costs, its greater energy production offsets this increase, resulting in comparable LCOE values. According to [31], the solar LCOE for small scale photovoltaic plants in Chile is approximately USD 57.1/MWh. The values obtained in this study are considerably higher than those reported in the reference. This difference is mainly attributed to the high investment costs combined with relatively low energy production for the level of investment required; in other words, due to an underutilisation of the projected charging infrastructure.

In the case of the LCOH, a value of 9.442 USD per kg of H₂ is obtained for the housing complex, while a value of 9.752 USD per kg of H₂ is estimated for the apartment complex. These values are relatively similar, consistent with the results observed for the LCOE. According to [31], the cost of producing, compressing and storing one kilogram of green hydrogen is 6.13 USD. As in the case of the LCOE, the results obtained from the economic analysis shows that the LCOH are higher for the studied case than those reported in the literature. This outcome can be explained by the high costs associated with H₂ production at this scale, combined with a production capacity that is not well matched to the level of investment, and a degree of underutilisation since ultra-fast charging does not demand a constant power.

Since the electricity cost has seen considerable changes in the past years, a final sensitivity study is performed considering the actual electricity price variations observed in the past five years (Nov 2020 to Jan 2026 [32]), as shown in Figure 9. The BT4.3 tariff was selected since it can deliver the power range needed for the considered EV chargers.

Figure 9.

Observed monthly electricity price for the BT4.3 tariff in Santiago

Based on the observed electricity price, a linear regression is used to deliver a price projection for the following 20 years, taking an average case and two variations: a pessimistic and optimistic case, by either adding or subtracting the standard deviation of the observations. The resulting NPV, calculated for each case study considering a loan, is compared to the base case in Table 9. The results with the projections are highly profitable compared to the original cases. The reason behind this difference is the high increase in electricity prices that has been observed in the past years in Chile, which ranges from 8% to 19% annually. While the result sounds promising, we caution that the local context does not assure that prices will continue to increase; since part of the increase was a calculation error that is expected to be partly reverted during 2026 [32]. However, it is still an interesting contrast to actually observe the dramatic effects of such an increase in the estimated profitability of this type of projects.

The resulting low competitiveness of H2-based solutions raises an important question. Are there other technologies that could be an alternative? The closest alternative are other storage systems, especially battery energy storage systems (BESS), which were not included in the present study. Even though BESS costs could be high, future work economic evaluation could be performed to evaluate its feasibility at the multifamily household residential scale. A high-power output would be desired for ultra-fast charging, which could lead to high investment costs; however, they have a high roundtrip efficiency (close to 90%) and therefore have lower conversion losses than H₂-based systems.

If, instead, EV charging does not have to comply with the ultra-fast standard, many other options may be more competitive, and even single household charging could be more convenient for housing complexes where parking spaces are close to electric outlets, especially by freeing users of charging time restrictions. However, the communal analysis would continue being a matter of innovation and research for apartment complexes due to the lack of space and lack of access to electricity in all parking spaces.

Another limitation of the present study is the overly deterministic demand model, which could greatly benefit from real-world data if available, or by including statistical deviations of the users’ expected behaviour. Still, this is an interesting area of innovation for EV charging startups and developers that are considering solutions for multifamily households. Representing and evaluating their proposals with a techno-economical basis could be a good way to accelerate the transition to sustainable transportation.