Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate

RESEARCH ARTICLE Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate Ursin Gsto¨ hl*, Stefan PfenningerID* Climate Policy Group, Institute for Environmental Decisions, ETH Zu¨ rich, Switzerland * ugstoehl@student.ethz.ch; stefan.pfenninger@usys.ethz.ch Abstract The idea that households produce and consume their own energy, that is, energy self-suffi- ciency at a very local level, captures the popular imagination and commands political sup- port across parts of Europe. This paper investigates the technical and economic feasibility of household energy self-sufficiency in Switzerland, which can be seen as representative for other regions with a temperate climate, by 2050. We compare sixteen cases that vary across four dimensions: household type, building type, electricity demand reduction, and passenger vehicle use patterns. We assume that photovoltaic (PV) electricity supplies all energy, which implies a complete shift away from fossil fuel based heating and internal com- bustion engine vehicles. Two energy storage technologies are considered: short-term stor- age in lithium-ion batteries and long-term storage with hydrogen, requiring an electrolyzer, storage tank, and a fuel cell for electricity conversion. We examine technological feasibility and total system costs for self-sufficient households compared to base cases that rely on fossil fuels and the existing power grid. PV efficiency and available rooftop/facade area are most critical with respect to the overall energy balance. Single-family dwellings with pro-

Accepted: December 17, 2019 Published: March 4, 2020 Copyright: © 2020 Gsto¨hl, Pfenninger. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the manuscript and its Supporting Information files. Funding: SP acknowledges funding from the Swiss Competence Center for Energy Research − Supply of Electricity (SCCER SoE), contract number 1155002546. http://sccer-soe.ch/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

at scales ranging from individual buildings to larger political regions [1]. However, maintain- ing a secure and affordable supply of energy remains an important aspect of the energy transi- tion. Energy self-sufficiency is primarily a political goal that requires technical solutions like long-term storage for its actual implementation, the feasibility and costs of which are still unclear [2]. Here we investigate the technical and economic feasibility of energy-independent households relying on PV electricity for the case of Switzerland, which can be seen as represen- tative for the situation in a temperate and highly industrialized country. The novelty of this study is that we use a single, integrated approach to investigate a range of building types and demand scenarios, while fully considering mobility and heating demand as part of the build- ing’s self-sufficiency. Switzerland is party to the Paris Agreement which has the main aim of limiting global tem- perature increase above pre-industrial levels to 2˚C [3]. The Swiss government has further- more adopted a net zero emissions target by 2050 [4]. Switzerland has also decided to phase out nuclear power following the Fukushima disaster, and the last nuclear power plant will have to shut down by 2034 [5]. In 2017, 31.7% of Swiss electricity came from nuclear plants, with the majority of the remainder from hydropower (59.6%) and only 4.7% from conventional thermal plants, and 4.0% from non-hydro renewables [6]. Since hydropower has only limited

additional potential, replacing the lost nuclear generation with other emissions-free electricity sources will be a major challenge, particularly as electricity supply rises along with electrifica- tion of heating and transportation [7]. Increasing decentralized production by households is an attractive option in this context: the residential PV potential has been estimated as between 11 to 19 TWh in Switzerland in 2050 [8]. If electric mobility becomes increasingly common, as we should assume will be the case, household electricity use will increasingly include charging of vehicles [9]. Thus, the question of decentralized and potentially self-sufficient electricity generation based on large-scale PV deployment combined with the large-scale deployment of electric mobility is particularly relevant, not just for Switzerland, but for any highly industrial- ized economy in a temperate climate facing a similar clean energy transition challenge. Energy independent residential dwellings have been addressed in the literature, as shown in the overview in Table 1. However, past studies were often constrained to special conditions, limiting their general applicability. Some of the studies outlined in the table help us build a comprehensive picture of how individual components in a self-sufficient household work, e.g. by optimizing combined PV and battery deployments [10–13]. Other studies analyze individ- ual cases of net zero-energy buildings [14, 15], but these studies do not include considerations

of electrified mobility and how that would affect building-level self-sufficiency. There are no studies with a view of the bigger picture that make more general statements about the potential for residential self-sufficiency in a country like Switzerland. The more systems-oriented studies are based on the concept of decentralized energy systems, where individual units are still con- nected to a local distribution network and the study aim is to design optimal local energy sys- tems (e.g. [12], [16]). In these systems, energy is supplied primarily through renewables and converted and stored in a range of different energy carriers. To synthesize past work into an overall understanding of the conditions under which household self-sufficiency based on PV electricity is possible, we build generic cases that vary on the dimensions of household type, buildings type, electricity demand, and electric mobility behavior, and build scenarios for their household energy demand by 2050. To construct these cases, we draw on data and make assumptions based on the detailed studies shown in Table 1. We investigate three questions: • In which of these cases, if any, are households able to satisfy their energy demand with self- produced electricity? Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 2 / 25

Table 1. Overview of key past studies on residential energy self-sufficiency [10–19]. Source Aim and methods Results Differences to our study GROSSPIETSCH, Thommes (10) Simulation model which shows how, when, and where a self-sufficient neighborhood is economically feasible. How: Through solar PV together with a battery system (1) or together with batteries and hydrogen storage (2). When: Costs have to decrease 21% (1) and 51% (2), respectively, to be competitive to households relying on fossil fuels. Where: Little seasonality (1) to high seasonality (2). • No timeframe • Ranging from low to high latitudes • Neighborhood containing a community of households of different sizes and an office building • No sensitivity analysis of factors • Technically not feasible with current efficiencies • No consideration of electric mobility Murray, Orehounig (11) The potential of long-term and short-term storage systems are modelled in a decentralized neighborhood with minimizing costs and CO2- emissions. Three scenarios are created and modelled until 2050. The model is tested on two sample municipalities in Switzerland. Modelled data regarding technologies and price developments until 2050. Development in the locations of Zernez and Altstetten, Switzerland, are presented. Furthermore, the importance of retrofitting the current building stock is mentioned because of the limitation of renewable energy. • No information about single buildings is given • Focus on the two case study locations • Local gas, electricity, and heating grid are present and used • No consideration of electric

mobility Zhang, Campana (12) Grid-connected PV-hydrogen/battery systems are investigated with the creation of three strategies for the hydrogen storage. Moreover, a high and a low- cost scenario are created. Strategies to produce and store hydrogen are dependent on the electricity price and make sense when a seasonal mismatch between load and PV production occurs. Hydrogen storage shows a better performance when grid power fluctuation is considered. • Connected to the grid • No clear relation to the residential sector • No information about sizing the hydrogen or battery system Weniger, Tjaden (13) Optimal sizing of PV-battery-systems in the residential sector are analyzed. In addition, a sensitivity analysis is conducted regarding PV and battery size in different cost scenarios. The absolute yield and the PV generator orientation will be of small relevance in the future. The most optimal PV sizes are reached in small-scale systems with high self-consumption rates and degree of self- sufficiency (above 70%). In a long-term perspective, PV systems with batteries will be the most economical solution. • Connected to the grid • Special case (single-family building) located in Lindenberg, Germany • Information about the household is restricted to its electricity demand of 4 MWh/year • No renewable heating system is considered • No hydrogen production is considered • Feed-in tariff is applied in the modelling • No consideration of electric mobility Good, Andresen (14) Solar energy solutions are modelled and compared with the aim of fulfilling the requirements of a net

zero energy balance. If the building only uses solar PV, the net zero energy balance shows the better performance than using solar thermal collectors in combination with PV or using the technology of hybrid photovoltaic-thermal modules. • Special case of a single-family building in Norway • Connected to the grid • No storage of electricity • No consideration of electric mobility Milan, Bojesen (15) A model is developed to optimally size a 100% renewable supply system while considering overall costs. It is applied on a case study in Denmark. The requirement of a net zero-energy building is successfully applied and optimally if the PV system is combined with a heat pump. Available rooftop area is not used for solar thermal collectors. • Connected to the grid • Special case with conditions of Northern Denmark • No hydrogen production is considered • No consideration of electric mobility Maroufmashat, Fowler (16) A generic mathematical model is developed to manage future communities which use hydrogen as an energy carrier. The model is applied to a case study consisting of four energy hubs which work together. Results are based on a case study located in Ontario, Canada. Annual total emissions and the levelized costs of hydrogen of the energy hub networks are presented. To produce hydrogen on-site is more economical than purchase it. The environmental advantages of hydrogen vehicles are assessed in addition. • Focus on hydrogen production only • A network of energy hubs is investigated with no further explanation of their composition • No detailed information about

the residential sector • No storage of electricity in batteries • Consideration of hydrogen fueled cars (Continued) Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 3 / 25

• Which factors influence the technological and economic feasibility of self-sufficiency? • How would energy costs for self-sufficient households compare to those still relying on fossil fuels and grid-based electricity by 2050? The more densely populated an area, the more difficult it will be to generate enough elec- tricity for that area with solar PV only. Single-family buildings therefore achieve self-suffi- ciency most easily. However, with technological improvement leading to more efficient PV panels, even some multi-family buildings can be self-sufficient. In all cases, cost driven by stor- age needs is a key issue, and implementing fully built-out rooftop PV on fully electrified build- ings while remaining connected to the grid is the most cost-attractive solution under almost all assumptions. Methods Definition of cases We define sixteen cases through which we examine energy self-sufficiency, aiming to represent expected state-of-the-art technologies available by 2050. The cases are constructed by varying four dimensions which each represent a binary choice on key assumptions for stylized housing types, as shown in Fig 1. Dimension 1 (household type) defines energy requirements of single-family households (SFHs) in single-family buildings (SFBs) and multi-family households (MFHs) in multi-family buildings (MFBs). Dimension 2 (building type) distinguishes between new and retrofitted buildings, with a focus on their differing energy requirements for space heating. Dimension 3 (electricity demand) is defined by behavioral changes and technology developments which

may lead to lower energy demand. Because future electricity demand cannot be predicted we differentiate between two stylized cases: a profound electricity reduction and a scenario with a conventional electricity use. This only affects electricity demand not already included in space heating and electric mobility (covered by dimensions 2 and 4, respectively). Dimension 4 assumes a complete electrification of passenger cars in both cases, with differentiation of mobility behavior based on a typical urban or rural household. This is an important dimension because driving behavior and developments in the car industry will profoundly influence elec- tricity demand [20–22]. We only consider battery electric vehicles (BEVs) here. Fuel cell elec- tric vehicles (FCEVs) can be seen as an alternative provider of electric mobility, but we do not Table 1. (Continued) Source Aim and methods Results Differences to our study Lang, Ammann (17) The creation of a model to simulate the performance of rooftop PV for different building types to assess the techno-economic potential of PV self- consumption. In the small residential sector, 40% of the produced electricity is self-consumed. In the large residential sector 80%. Furthermore, the internal rate of return is calculated to measure profitability. • The aim is not to be self- sufficient • No facade PV • Connected to the grid • No storage of electricity Prognos (18) The aim is to build a quantitative basis for political and societal discussions. This is done by modelling three scenarios to investigate the effect of framework

conditions, influencing factors, political interventions, and climate and energy goals. Quantitative data of the three scenarios until 2050. • Modelling of the entire Swiss energy landscape • No statement regarding self- sufficiency • Diverse technologies are modelled SFOE (19) Different heating systems are simulated with the aim to find the most efficient combination of heat pumps and solar energy solutions. Different advantages and disadvantages are mentioned. The most economical system is composed of an air-to-water heat pump and solar PV. • Special case of a single-family building with a fixed energy consumption • Only heating system is of interest • No storage of electricity https://doi.org/10.1371/journal.pone.0227368.t001 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 4 / 25

explicitly consider them, instead assuming that their usage patterns and resulting electricity demand would be similar as those for BEVs. Table 2 summarizes all assumptions used to quantify the four dimensions. Many of the future trends are taken from the “New Energy Policy” (NEP) scenario from the Swiss Energy Perspectives 2050 created on behalf of the Swiss Federal Office of Energy (SFOE) by Prognos (18). Each of the cases is modelled for a self-sufficient and “alternative” future. The self-sufficient cases assume full electrification of mobility and heating demand. The alternative cases assume that household energy is drawn from the electricity and gas grids, as well as the continued use of internal combustion engine (ICE) vehicles. We assume an equal purchasing price and life- time of BEVs and ICEs by 2050, and the cost of vehicle purchase is not further included in our cost calculations. The use of EV batteries for additional balancing in households is not consid- ered given the uncertainty about EV availability based on different mobility profiles. To com- pare the cases of fully self-sufficient electrification with a grid-connected full electrification scenario, we also investigate a third future where buildings remain connected to the grid. These cases are designed with or without on-site battery storage (in battery cases, all of the electricity is assumed to pass the battery), and with rooftop PV efficiencies of 22.1% and 27.2%. Two market scenarios are used. Under a scarce market, electricity purchasing price is

0.31 CHF/kWh (+20% of the 2050 assumption from [11]) and a selling price half this purchas- ing price. In an abundant market the purchasing price is of 0.21 CHF/kWh and no electricity can be sold back to the grid. All PV oversupply is sold, and in the no-battery case, all demand has to be satisfied at market prices. This third future is used solely to compare the cost of grid- connected electrification with fully self-sufficient electrification. The key assumptions relate to the area available for PV systems, demand which needs to be met by this PV capacity, and the availability of storage. PV potential. To determine the available solar electricity production, each case was opti- mized for an optimal solar PV area to meet its demand, by applying a generalized reduced gra- dient descent algorithm as implemented in the Microsoft Excel Solver add-in (see S2 and S3 Data). The monthly electricity balance was calculated for each case and PV cell efficiency by Fig 1. Outline definition of the sixteen self-sufficient cases with abbreviations. D1 (green) represents the first dimension of household types, D2 (blue) represents the second dimension of building types, D3 (orange) represents the third dimension of projected electricity demand, D4 (yellow) represents the fourth dimension of BEVs and spatial behavior. Each case is coded with an abbreviation. First letter: S = SFB, M = MFB; Second letter: N = New building, R = Retrofitted building; Third letter: P = Profound electricity reduction, C = Conventional electricity use; Fourth letter: R = Rural area, U = Urban area.

https://doi.org/10.1371/journal.pone.0227368.g001 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 5 / 25

subtracting the electricity demand from the produced electricity. The algorithm computes the optimal area of solar PV, considering a surplus production for hydrogen storage, and by taking restrictions on the rooftop and facade sizes into account. An optimal solar PV area can be reached when the energy balance is exactly equilibrated at zero. The objective is to produce a sufficient amount of energy with the given parameters of Table 2 while minimizing the panel Table 2. Assumptions for all cases. Dimension 1 determines the dimensioning of SFBs and MFBs with their electricity consumption and domestic hot water (DHW) demand. Dimension 2 determines usable rooftop share and yearly space heating energy demand of newly built SFHs and MFHs today and in 2050 with the individual reduction potential and the projected SPF of air-to-water heat pumps in 2050. Dimension 3 determines changing electricity demand with the consideration of cooling and lighting demand, and dimension 4 considers the required electricity for transportation, per house unit and month, derived by estimating in passenger kilometers per per- son in both a rural and urban setting. All cases Overall assumptions Sources Inverter efficiency 99% [23] Performance ratio 0.9 [24] Solar irradiation, rooftop 1,000 kWh/m2/year [25] Solar irradiation, facade 800 kWh/m2/year [10], [26], [25] Dimension 1: building type SFBs MFBs Occupancy rate per household 4 4 [10], [27] Occupancy rate per building 4 20 [10] Stories 3 5 [10], own assumptions Floor area 200 m2 900 m2 [10], [28], [29], [30], [31]

Rooftop slope 35˚ 35˚ [32], [33] Rooftop area 81.4 m2 219.7 m2 Own assumptions Facade area for PV production (33% of one facade side) 25.5 m2 67.7 m2 Own assumptions Current electricity consumption of each house unit 5,200 kWh/year 19,250 kWh/year [34] DHW demand in 2050 10.9 kWh/m2; -26% compared to 2020 [18] 12.6 kWh/m2; -26% compared to 2020 [18] [14], [19], [29], [30], [31], [35] DHW efficiency factor in 2050 3.5; +22% compared to 2020 [18] 3.5; +22% compared to 2020 [18] [18], [36] Dimension 2: building age New buildings Retrofitted buildings Usable rooftop area 100% 70% [37] Yearly energy demand of SFHs–now 16.0 kWh/m2 22.5 kWh/m2 [14], [19], [29], [30], [31], [35], [38] Reduction potential -53% -56% [18] Yearly energy demand of SFHs–in 2050 7.5 kWh/m2 9.9 kWh/m2 [14], [18], [19], [29], [30], [31], [35], [38] Yearly energy demand of MFHs–now 18.2 kWh/m2 20.9 kWh/m2 [14], [19], [29], [30], [31], [35], [38] Reduction Potential -34% -39% [18] Yearly energy demand of MFHs–in 2050 12.0 kWh/m2 12.7 kWh/m2 [14], [18], [19], [29], [30], [31], [35], [38] SPF of air-to-water heat pumps–in 2050 4.8; +46% compared to 2018 [18] 4.4; -9% compared to new buildings [39] [18], [36], [39], [40], [41] Dimension 3: demand Profound reduction Conventional use Decrease of electricity reduction compared to [34] -29% -7% [18] Climate change impact on cooling electricity demand +2.5% of today’s electricity consumption [34] +2.5% of today’s electricity consumption [34] [42], [43] Share of lighting of the entire electricity demand 13.3% 13.3% [38]

Dimension 4: mobility Urban SFBs Rural SFBs Urban MFBs Rural MFBs Passenger km (pkm) in 2050 per person times occupancy rate 4,814 7,554 4,201 6,439 [44], [45] Pkm in 2050 per house unit 19,255 30,215 84,021 128,785 [45], [46] Required kWh/100 km in 2050 12 12 12 12 [47], [48], [49] Required kWh in 2050 per house unit 2,311 3,626 10,082 15,454 Calculation based on above sources Required kWh in 2050 per house unit and month 193 302 840 1,288 Calculation based on above sources https://doi.org/10.1371/journal.pone.0227368.t002 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 6 / 25

area, which reduces costs. Additionally, an increase in PV efficiency enables a smaller area to cover for the same output. Rooftop and facade PV electricity production is unevenly distrib- uted over the year. Average monthly production was determined through the online calculator tool of EnergieSchweiz (33) for ten locations in Switzerland (1072 Forel, 2942 Alle, 3076 Worb, 3930 Visp, 5076 Bo¨zen, 6020 Emmen, 6952 Canobbio, 7000 Chur, 8046 Altstetten, 8853 Lachen). The tool uses long-term solar irradiance data for the specific location, as well as installed module capacity, tilt and azimuth angles, to compute long-term average monthly electricity generation. Panels were assumed to be fixed with a tilt angle according to the roof- top slope given in Table 2 and receive a monthly irradiance share calculated based on the aver- age of a south-oriented, west-oriented, east-oriented panel. The same approach was used for facade PV systems. Data for the feasibility assessment with a realistic generation profile from Renewables.ninja are described further below. Demand. The energy demand for each case is quantified on a monthly basis to capture the fact that lighting and space heating demand vary seasonally [38]. Regarding dimension 1, the household types need a specific size, which we base on adjusted data from measurements performed by Grosspietsch, Thommes (10). Annual energy demand of SFHs and MFBs for domestic hot water (DHW) was calculated by taking the mean of reported future demand

from the sources given in Table 2, in order to arrive at a robust estimate of potential demand reductions. Space heating demand is not only influenced by the household type but also by the building type. The sources given in the table were used to derive an average value for new and retrofitted SFBs and MFBs [29–31]. We assume an air-to-water heat pump for space heating based on assessing different advantages and disadvantages of several possible heating systems [18, 19, 40, 50]. As for mobility demand, we assume that buildings in an urban setting require less vehicle range than buildings in a rural setting. This is based on data of the Swiss Federal Statistical Office (FSO) [45] and the Swiss Federal Office for Spatial Development [44, 46]. We also assume an occupancy rate in cars of 1.6 in 2050 based on the fluctuating trend of the past described by the [45]. Storage. In periods where PV-generated electricity cannot cover demand, residents will have to rely on stored energy. Short-term storage can be covered by lithium-ion batteries. Lor- enzi and Silva [51] state that efficiencies of 92% were already achieved in 2016. Hence, we assume 92% to be a solid value also for 2050. Moreover, the discharge depth is almost 100% and the monthly self-discharge rate is estimated between 1% and 5% [52, 53]. Therefore, we assume the discharge depth and the monthly self-discharge rate to be negligible in 2050, in particular given that the systems are in use on a daily basis. A further assumption is that all

produced energy has to pass the battery system. Battery dimensioning is assessed for each case separately with a battery storage capacity twice as large as the average daily energy demand in kWh. This size is chosen as Tesla [54] suggests that a battery should be able to cover one full day of energy usage, while Umweltarena [55] recommends a battery which covers up three days. Long-term storage is based on hydrogen production and reconversion into electricity. The processes of electrolysis, compression and fuel cell usage require more steps than electric- ity storage in batteries. PEM electrolyzers have an efficiency of 70% [56] in 2017 and projec- tions reach efficiencies of up to 85% [57] or higher [58] in 2030. Compression is assumed to be integrated in the electrolyzer with a pressure of 30 bar [59]. An electrolyzer efficiency of 80% is thus reasonable in 2050. The efficiency of PEM fuel cells is provided by various sources [10, 49, 60]. Based on positive efficiency estimations by Miotti, Hofer (49) of 52% for 2030, we assume this value to be reached in 2050. We assume that electricity produced in the fuel cell also passes through the battery system. All of these steps result in an overall long-term storage efficiency of 38.3%. For each case, the size of the hydrogen tank is dimensioned to guarantee sufficient storage for hydrogen with an energy content equivalent to 100% of the building’s yearly end- Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate

PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 7 / 25

use demand of electricity. The term end-use demand is defined as the energy demanded by users and should not be mixed up with the term final energy consumption which represents the amount of energy input needed to reach the end-use demand [38]. Costs If the self-sufficient cases are cost-competitive compared to similar cases still relying on fossil fuels and electricity from the grid, they could be attractive investments for households. To examine whether this might be the case, we calculate total discounted costs for each case: Total costs CHF kWh � � ¼ PT t¼1 investment costst þ O&M costst ð1 þ iÞt ð1Þ Variable T indicates the observed time horizon in both equations. t = 1 is the year 2050. We assume a time horizon of 20 years and set the discount rate i to 4% in 2050, which represents the contemporary average discount rate of residential consumers [61]. Eq 4 is based on the LCOE definition of Hernandez-Moro and Martinez-Duart [62] and comprises investment costs (capital costs and balance-of-system (BOS) costs) and operation and maintenance (O&M) costs. BOS costs are defined as all costs occurring in the purchasing period minus the capital costs of a product. The BOS costs of a PV plant include, for instance, the costs of wiring, the inverter, planning, installation, and the transformer. O&M costs occur every year because of services to keep the technology in operation. Efficiency improvements of electrolyzers and fuel cells are included in the equations in Table 3 by assuming a cost reduction of 33.3% for

electrolyzers and of 66.6% for fuel cells by 2050. The capacity of electrolyzers and fuel cells are adjusted to data of a multi-family zero-energy building in Bru¨tten, Zurich [63], with 4.8 kW in SFBs and 9.7 kW in MFBs for electrolyzers, and 2.1 kW in SFBs and 4.1 kW in MFBs for the fuel cells. Furthermore, batteries have to be replaced after operating for 15 years. Cost data used for the self-sufficient and fossil/grid alternative cases are shown in Table 3 and Table 4, respectively. Where necessary, currencies are adapted to inflation and converted to 2019 CHF. Sensitivity analyses To test the feasibility of the self-sufficient cases in a more realistic environment, we consider daily data of a real location in addition to average monthly data. We assess the average of a south-oriented, an east-oriented, and a west-oriented rooftop in Worb, a municipality located in the Swiss Plateau region, for 2000 to 2015 using the PV simulation model provided by Renewables.ninja [74]. Renewables.ninja uses the GSEE simulation software [75], and was set up to compute hourly solar electricity generation based on module capacity, tilt and azimuth angle, and bias-corrected satellite-derived irradiance data from CMSAF SARAH, then averag- ing hourly to daily values. This location was chosen because of its position in the Swiss Plateau where most of the Swiss population is domiciled [76]. In a first step, Renewables.ninja pro- vided the average daily output for the capacity of one kWp installed capacity of solar PV on a

specific day (Q). Running a simulation for south-oriented (Qsouth), west oriented (Qwest), and east-oriented (Qeast) rooftops allows computation of the average daily capacity with one kWp of solar PV installed. To test the robustness of the cases in a generic environment, we normal- ize the yield data from Worb to 1,000 kWh/m2 by applying a 10.8% reduction to its 1,122 kWh/m2 annual average. The risk of not being able to provide enough energy is hedged by enforcing a hydrogen energy overproduction 1.75 times the found optimal value of produc- tion. The individual capacities per case (LCaseX) are calculated based on the knowledge of required needed rooftop and facade area. The output (CCaseX) in kWh/day is this performance Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 8 / 25

multiplied by 24 hours. Eq 1 shows the entire approach: CcaseX kWh day � � ¼ ðQsouth þ Qwest þ QeastÞ 3 � kW kWp � � � 1 0:108 ð Þ � LCaseX ½kWp� � 24 h day � � ð2Þ The necessary yield to cover energy demand on each day (DCaseX) is computed by dividing the optimized monthly energy production (YMonthZ), which is required to satisfy the demand in this month, through the amount of days of this particular month (NMonthZ): DCaseX ¼ YMonthZ NMonthZ ð3Þ On days where the difference between DCaseX and CCaseX is negative, the shortfall has to be compensated by energy from the hydrogen system. We assume that if the yearly positive off- sets, with the inclusion of the long-term storage efficiency in the hydrogen system of 38.3%, are high enough to outbalance the negative offsets, it is possible to rely fully on the energy pro- duced in this year. Thus, the following inequality must hold true for a case to be considered Table 3. Cost assumptions for the self-sufficient cases. All costs are in 2019 CHF. Price in 2050 [CHF2019] Source Lifetime [years] Source Solar PV Investment costs [CHF/m2] 200.- [11], [64], own assumptions 25 [11] Investment costs [CHF/kWp] for 22.1% efficiency 905.- - - - Investment costs [CHF/kWp] for 27.2% efficiency 735.- - - - O&M costs [CHF/kWh] 0.02 [11], [65], own assumptions - - Heat pump Investment costs–SFB 36,000.- [66] 20 [11] Investment costs–MFB 136,800.- Own assumptions - - O&M costs 200.- [66] - - Battery Capital costs [CHF/kWh] 265.- [48], [54], [67], own assumptions 15 [53], [54], [61], own assumptions

BOS costs 2000.- [68] - - Additional Hardware costs (part of investment cost) 740.- [54] 15 [53], [54], [61], own assumptions O&M costs 4.5% of investment costs [61] - - Electrolyzer Capital costs [CHF/kW] 70 633 CHF ½ � � X 0:276 kW ½ � � 2 3 [69], [70], own assumptions 22 [11], own assumptions BOS costs 30% of capital costs [61] - - O&M costs 5% of capital costs [61] - - Storage tank Investment costs 600 CHF m3 � � � XCaseX m3 ½ � [71], own assumptions 22 [11] BOS costs 2500 Own assumptions - - O&M costs 3% of investment costs [61] - - Fuel cells Capital costs [CHF/kW] 60 506 CHF ½ � � Y 0:177 kW ½ � � 1 3 [11], [61], own assumptions 22 [11], own assumptions BOS costs 30% of capital costs [61] - - O&M costs 10% of capital costs [61] - - https://doi.org/10.1371/journal.pone.0227368.t003 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 9 / 25

self-sufficient with a realistic PV generation profile: PDneg � PDpos � 38:3% ð4Þ Where Dneg are days where DCaseX > CCaseX and Dpos are days where DCaseX < CCaseX. We also examine the sensitivity of results to both technical and cost uncertainties. The effi- ciency of PV modules is a key parameter due to the exclusive reliance on PV to generate elec- tricity. We assume that our rooftop PV systems are sc-Si modules. Other work has shown that an efficiency of 27.2% can be reached by 2050 with a yearly increase by 0.3% [24]. We examine a range of PV efficiencies from the current efficiency of 17% to 27.2% [24, 25], focusing in par- ticular on two cases: 22.1% (assuming an annual increase by 0.15% until 2050) and 27.2% (annual increase by 0.3%). According to Mu¨ller, Folini [77], the average yearly increase in effi- ciency of sc-Si cells over the last decade was 0.17%, so our assumptions are within a realistic range. We assume that CdTe thin-film modules are used for facade electricity production as they are more suited for difficult light conditions or when PV modules should be integrated into building parts with a heterogenous structure [23]. Efficiency of CdTe thin-film cells are assumed to be 0.787 times that of sc-Si cells, based on their relative efficiencies under standard conditions from [78]. Other technical parameters as detailed in the results below are varied while PV efficiency is kept at 27.2%. A final component of our sensitivity analyses are the cost of key components, to examine their effect on the economic feasibility of the self-sufficient

cases. This includes costs in the alternative (fossil fuel and grid connected) cases. Their total costs are strongly influenced by the energy prices for electricity, gas, and gasoline. Finally, dis- count rate also affects total costs, so we vary it from its default of 4% to 0% and 6%. Results Feasibility of self-sufficiency The composition of the end-use demand for electricity we use based on the assumptions laid out above is shown in Fig 2. It consists of domestic hot water (DHW), space heating, general Table 4. Cost assumptions for the alternative cases exchanging energy with the grid and using fossil fuels. All costs are in 2019 CHF. Price in 2050 [CHF2019] Source Lifetime [years] Source Gas heating system New installation—SFH 20,000.- [72] 20 [11] Replacement—SFH 14,000.- [72] 20 [11] New installation—MFH 76,000.- [72], own assumptions 20 [11] Replacement—MFH 70,000.- [72], own assumptions 20 [11] O&M costs [CHF/year] 650.- [72] - - Air conditioning Price–SFB 6000.- [73] 20 Own Assumptions Price–MFB 10,000.- [73] 20 Own Assumptions O&M costs [CHF/system] 150.- [73] - - Energy prices Change 2020–2050 Electricity price [CHF/kWh] 0.262 [11] +24% [18] Gas price [CHF/kWh] 0.175 [18] +43% [18] Gasoline price� [CHF/kWh] 0.29 [18] +29% [18] � with a density of 0.775 kg/l and a heating value of 11.6 kWh/kg https://doi.org/10.1371/journal.pone.0227368.t004 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 10 / 25

electricity demand (for lighting, cooking, devices, etc.), and electricity for charging electric vehicles. Based on this demand, and using the assumptions for technology deployment outlined in the methods above, Fig 3 shows the net energy balance as a function of the chosen scenario and the assumed PV module efficiency. All cases except for those shaded in dark red (marked 4 in the figure) can supply enough electricity over the entire year to meet annual demand and can thus be considered net zero-energy buildings. The net zero energy building boundary (dark red line) indicates this. In achieving net zero energy, only cases with MFBs show trouble at low PV efficiency values, but all cases are successful with high PV efficiencies. MFBs gener- ally have more difficulties than SFBs because of their smaller available area for solar PV in rela- tion to demand. If we add the condition that buildings must be able to supply additional electricity for hydrogen storage to use in net-negative months, more cases fail, indicated by the light red shading (marked 3). If we further tighten the conditions to demand hydrogen pro- duction be 1.75 times the amount determined to supply net-negative months in order to hedge against low solar yield for several days in a row, more cases fail, indicated by light blue shading (marked 2). This leaves the area shaded in dark blue (marked 1 in the figure): cases which are able to supply total annual demand as well as sufficient hydrogen production to balance net- negative months and shorter-term variability. These cases can thus be considered true zero

energy buildings that could supply all of their inhabitants’ end-use electricity demand in a fully self-sufficient manner. The dark blue line marks this true zero energy building boundary in the figure. Fig 4 focuses on the feasibility of true zero energy buildings (i.e., our condition that 1.75 times the optimal amount of hydrogen can be produced). It shows how far above or below the balanced (supply equals demand) line the individual cases are, as a function of PV efficiency. Installed PV system capacities range for cases with SFBs from 12.7 kWp (SNPU, 27.2% rooftop PV efficiency) to 18.1 kWp (SRCR, 22.1% rooftop PV efficiency) while cases with MFBs require a capacity between 55.2 kWp (MNPU, 27.2% rooftop PV efficiency) and 77.4 kWp (MRCR, 22.1% rooftop PV efficiency). Because most cases include the use of facade PV, and the monthly share of solar energy production differs between rooftop PV and facade PV, installed capacities are not exactly equal for 22.1% and 27.2% efficiency. The discussion of PV system capacities leads us to the consideration of overall system dimensioning. Fig 2. End-use electricity demand of each case in MWh/year. Retrofitted buildings require more energy than new buildings; this difference is indicated with the black bar and percentage difference on top of the total end-use electricity demand. https://doi.org/10.1371/journal.pone.0227368.g002 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 11 / 25

System dimensioning for self-sufficiency The battery, hydrogen, and PV system dimensions required for full self-sufficiency (true zero energy buildings) and assuming 27.2% rooftop PV efficiency are given in Table 5. Since the fuel cell and electrolyzer sizes are fixed (4.8 kW in SFBs and 9.7 kW in MFBs for electrolyzers, 2.1 kW in SFBs and 4.1 kW in MFBs for fuel cells) they are not shown in this table. The bottom four multi-family cases marked with a � are those which consider our design limits for allowed PV area (i.e., they are outside the true zero energy boundary in Fig 3). The table illustrates that the key factor causing lower feasibility of self-sufficiency for both retrofitted houses and multi-family houses is that of limited rooftop space for PV, relative to demand: the SRCR case is the only single-family case that needs to add facade PV to fulfill its design requirements. While we also assume that the additional demand in retrofitted buildings is higher than that in new buildings, which makes these cases more difficult to begin with, it is this area limitation which is the key design parameter. Fig 3. Feasibility of achieving true zero energy and net zero energy requirements as quantified by the energy balance level achieved, and as a function of the scenario and rooftop PV module efficiency. Only the last two identifying letters for each scenario are shown for increased readability, with New/Retrofitted and Single/ Multi-family building given by the lines spanning the x-axis below the scenario labels. https://doi.org/10.1371/journal.pone.0227368.g003

Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 12 / 25

We explore this in more detail in Fig 5, showing the relative share of the area we assume to be available used in each case as a function of scenario, and comparing 22.1% and 27.2% roof- top PV efficiencies. Rooftop area is always used first because of the higher efficiencies com- pared to facade PV. With an efficiency of 22.1%, the available facade area in SFB cases is sufficient, except for case SRCR, which can also be seen in Fig 3 where it is the only case out- side the true zero energy building boundary at 22.1% efficiency. Where cases reach beyond the grey shaded region, they fail to fulfill our design requirements. However, they are not at their Fig 4. Cases with SFBs (left) and MFBs (right) considering backup hydrogen production in the size 1.75 times the optimized amount in net negative months. Values above zero net energy availability indicate cases that are successfully true zero energy buildings throughout the year. Values below this line indicate failed cases. The PV module efficiency is for rooftop PV systems; facade PV system efficiency tracks that of rooftop PV efficiency as indicated in the methods section above. https://doi.org/10.1371/journal.pone.0227368.g004 Table 5. Battery and hydrogen tank sizes as well as area covered by the PV system per case. Required rooftop and facade area are given to fulfil the energy demand with the assumption of a rooftop PV efficiency of 27.2% and facade PV efficiency. Cases marked with � are considered to exceed our limits for PV area and thus not able to be fully self-sufficient.

Battery size (kWh) Hydrogen tank size (m3) Rooftop PV area (m2) Facade PV area (m2) SNPU 38.6 9.0 46.7 0 SNPR 45.8 10.7 55.1 0 SNCU 44.9 10.4 54.2 0 SNCR 52.1 12.1 62.7 0 SRPU 39.4 9.2 48.2 0 SRPR 46.6 10.8 56.6 0 SRCU 45.7 10.6 55.7 0 SRCR 52.9 12.3 57.0 10.3 MNPU 162.5 37.8 200.0 0 MNPR 191.9 44.7 219.7 21.1 MNCU 185.6 43.2 219.7 11.7 MNCR 215.0 50.0 219.7 61.1 MRPU� 164.5 38.3 153.8 71.9 MRPR� 193.9 45.2 153.8 121.3 MRCU� 187.6 43.7 153.8 112.0 MRCR� 217.0 50.5 153.8 161.3 https://doi.org/10.1371/journal.pone.0227368.t005 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 13 / 25

absolute maximum since we assume that only one third of one facade side is used for PV. Our results for the MFB cases require up to 3.7 times our designated facade area. Our facade use could be considered a conservative estimate. In most settings, however, substantially exceeding it is likely implausible, given buildings that were not designed for facade PV, and since we assume that average solar irradiation on all sides of the facade is 800 kWh/year. Nevertheless, it remains theoretically possible to consider two thirds of the sunniest facade side and one third each of less sunny facade sides covered; this would represent a relative use of 4 and would make true self-sufficiency technically possible in all of our cases. Costs for self-sufficient and alternative cases The total discounted cost of the different components in each case is shown in Fig 6 for both self-sufficient and alternative cases. This cost data shown in the figure are for a rooftop PV effi- ciency of 22.1%. Rooftop and facade PV would get a marginally smaller cost if plotting 27.2%; but as the figure clearly shows, they are not the major cost components. The operation and maintenance costs in the self-sufficient cases are large; and on the same order as the invest- ment cost for the PV system. These maintenance costs are mainly for the storage components (Table 3). Alternative cases assume no electrification of transport or heat; heating comes from gas burners, cooling comes from air conditioners (with electricity drawn from the grid), and

transportation is by internal combustion engine vehicles. The fuel cost shown in the figure combines the cost of gas for heating and the cost of gasoline for driving. As a comparison, we show (in Fig 6C) the fuel cost change that would be required for each alternative case to become cost-equivalent with the corresponding self-sufficient case, all other assumptions remaining equal. A 100% required fuel cost change means fuel costs would need to double. Cases with MFBs require a lower percentage increase of the prices than cases with SFBs, as the cost of the corresponding alternative cases have higher shares of energy prices. Another reason is that investment costs of technologies in MFBs are distributed across more useful energy demand met than in cases with SFBs. An important point that is not visible in Fig 6 is that self-sufficient cases require less energy in terms of their final energy consumption than alternative cases. Heat pumps and EVs are more efficient than gas heating systems and Fig 5. Relative use of rooftop and facade area per case with a rooftop PV efficiency of 22.1% and 27.2%, for (a) SFBs and (b) MFBs. The shaded 100% use border signalizes a full occupancy of the total available rooftop area. The shaded 70% use border signalizes full occupancy of the available rooftop area in retrofitted buildings, given that we assume those buildings have more pre-existing roof infrastructure which prevents 100% use. https://doi.org/10.1371/journal.pone.0227368.g005

Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 14 / 25

ICEs. Fig 7 shows this difference in final energy consumption of alternative and self-sufficient cases per person. Total costs in the self-sufficient cases are generally higher than in alternative cases. How- ever, as becomes clear in Fig 8, grid-connected fully electrified alternatives are almost always cheaper than fossil fuel alternative cases. In other words, unless electricity costs develop Fig 6. Share of total cost across different system components for (a) self-sufficient and (b) alternative cases, as well as (c) the percentage increase in fuel cost that would be required for alternative cases to become cost-equivalent with self-sufficient cases, assuming no other changes in component costs. All of these cost figures show total discounted costs over the entire lifetime of the systems, which is assumed to be 20 years. https://doi.org/10.1371/journal.pone.0227368.g006 Energy self-sufficient households with photovoltaics and electric vehicles are feasible in temperate climate PLOS ONE | https://doi.org/10.1371/journal.pone.0227368 March 4, 2020 15 / 25