Electric Road Systems: A No-Regret Investment with Policy Support

doi:10.21203/rs.3.rs-3372572/v1

Download PDF

Article

Electric Road Systems: A No-Regret Investment with Policy Support

https://doi.org/10.21203/rs.3.rs-3372572/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This study addresses the pressing issue of reducing greenhouse gas (GHG) emissions in the road freight sector. We investigate whether Electric Road Systems (ERS) for heavy duty trucks are a viable solution for the European Union to align the sector with economy-wide GHG reduction targets. Results are based on simulations in MOSTACHI, a new model for studying interaction effects between competing charging infrastructure and cost-minimizing vehicle operators. ERS impact is assessed on transport cost, GHG emissions, resource demand, and competing charging infrastructure. We find that for ERS to be a no-regret investment, an accompanying policy is needed that penalizes non-participation. This ensures the economies of scale necessary for the ERS infrastructure’s viability and reduces transport cost on average. With policy, ERS increases battery-electric truck uptake and GHG reductions and reduces demand for batteries and fossil and renewable fuels, throughout the infrastructure lifetime.

Scientific community and society/Energy and society/Political economy of energy

Scientific community and society/Energy and society/Energy policy

Earth and environmental sciences/Environmental social sciences/Climate-change mitigation

Physical sciences/Energy science and technology/Energy infrastructure/Power distribution

The European Scientific Advisory Board on Climate Change (ESABCC) recommends reducing EU economy-wide greenhouse gas (GHG) emissions by 70% by 2030 and 90–95% by 2040, relative to 1990¹. This recommendation is significantly more aggressive than current targets of 55% reduction by 2030 and net zero by 2050². In contrast to other sectors, transport emissions are still increasing and alone risk using up the EU’s entire carbon budget³. Heavy-duty vehicles (HDVs) are a particularly challenging segment, illustrated by that proposed targets for decarbonization would merely realize a 64% GHG reduction by 2050⁴. New solutions are needed to enable progress in line with the rest of the economy.

This study contributes to broader research into whether Electric Road Systems (ERS) should be installed along major European transport routes to accelerate decarbonization of HDV traffic. In contrast to static charging technologies, ERS technologies allow dynamic charging, i.e., continuous power transfer for propulsion and battery charging while driving. Various conductive and inductive ERS technologies are under evaluation in several EU countries and globally, with no commercial installations to date.

Cost is the primary factor driving selection of freight solutions^5,6 and prior research has demonstrated that ERS can bring systemic cost savings. This is because destinations can be reached with smaller battery packs than if electric vehicles are charged statically, reducing vehicle cost^7–11 and raising cargo capacity. However, due to strong economies of scale and scope^8,9, the infrastructure itself can only be cost competitive at high participation by traffic, which further requires extensive – likely international – ERS buildout.

Sensitivity analysis has revealed that high ERS participation is not guaranteed, as future cheap depot charging may incentivize defection, triggering a vicious circle of increasing costs and further defection¹². If fears of stranded assets cannot be settled, ERS is unlikely to play a significant role in transport decarbonization.

This study improves on prior work by simulating previously identified but unexplored competition effects and feedback loops in time and space between HDVs, batteries, ERS and static charging infrastructure. Our main contribution is to test if policy that breaks the negative feedback loop by guaranteeing high ERS utilization would ensure ERS has long-term economic and environmental net positive effects even under ERS-adverse conditions. This would make ERS a no-regret option, greatly increasing the probability that the infrastructure gets built. Findings are contextualized and explained by further quantifying return on investment and ERS impact on demand for batteries and static charging. Software is provided for transparency and to enable replication of the analysis for other regions.

Limitations

Some ERS technologies support charging of both light and heavy vehicles. Though out of scope, prior research suggests economic savings potential from ERS may be greater for light vehicles^13,14, as these represent the majority of total expected battery demand from road traffic.

This study focuses solely on the economic aspects of ERS, assuming reliable and standardized technology at no more than twice expected cost. To date, no ERS technology is sufficiently mature for large-scale deployment. Unanswered questions remain regarding road durability impact, long-term maintenance costs, cybersecurity, supply chain capacity and electromagnetic compatibility^15,16. Wide-spread adoption of ERS also requires further standardization for international interoperability and integration into regulatory frameworks (e.g., the EU Alternative Fuels Infrastructure Regulation)^17,18.

While cost is an important driver, political, social and behavioral factors do influence the adoption of new technology beyond what is captured in this simulation study. These factors should be given greater attention in future work that looks at what policy instruments would achieve the desired effects with minimal adverse effects.

This paper is partly a generalization to European conditions of results in a technical report analyzing ERS in Sweden¹⁹. It therefore uses only Swedish road and transport data, which may introduce biases. Key model parameters have been adjusted where Sweden notably diverges from the EU average: biofuel supply cap (reduced from 35% of total internal combustion engine (ICE) fuel consumption in 2020 to 5%), electricity price (30% lower in Sweden) and GHG intensity (65% lower in Sweden in 2020), and CO₂ valuation (reduced from 700 €/ton to 200 €/ton). The study aligns its GHG emissions forecasts with EU reduction targets, not Swedish.

By international comparison, Sweden exhibits relatively sparse traffic and low potential for solar power. Annual average daily traffic (AADT) by heavy vehicles on the core Swedish road network is 1,000–5,000, while main roads in central Europe are in the 6,000–18,000 range²⁰. ERS economics improve with denser traffic. The shift to daytime charging brought by ERS (see results) may be a desirable property in electricity grids where solar power is a major electricity source. Such differences are not compensated for and contribute to the ERS-adverse conditions that are the focus of this study.

Limitations in the employed simulation model and input data are discussed in Methods.

Simulation Model

This paper introduces MOSTACHI (Model for Optimization and Simulation of Traffic And CHarging Infrastructure), a free and open-source agent-based simulation tool designed to study interaction effects in time and space between logistics patterns, competing charging infrastructure and cost-minimizing vehicle operators. The model capabilities and architecture are shown in Fig. 1 – for details, see Methods. MOSTACHI is. Simulation takes place consecutively over seven five-year periods, 2020–2054.

As MOSTACHI is a cost minimizing simulation tool, hydrogen fuel cell trucks are excluded due to expectations of levelized costs higher than for internal combustion engine trucks (ICETs)^21–25, which are included. Plug-in or ERS hybrids may have a role to play as a transitional technology, but the only zero-exhaust emission option in the simulation is battery electric trucks (BETs), mainly to reduce the computational complexity.

Calibrated transport pattern data²⁶ for Sweden from the nationally well-established Samgods model²⁷ and routed using Open Street Map. The data represent an origin-destination (OD) matrix of total annual goods tonnage transported, and total number of trips in 2017, between approximately 200,000 pairs of regions inside and outside of Sweden, for four truck weight classes 3.5–60 tons. Equivalent data exist for other countries.

Generalization of the simulation to a European context is made possible by adjusting model parameters (electricity prices and GHG emissions, GHG valuation, biofuel supply cap, etc.). No subsidies are assumed anywhere, except for fossil GHG emissions which are not taxed at their full valuation. Values for each five-year period for the approximately 200 global model parameters are provided as supplemental material.

Scenario-Based Robustness Analysis

Five scenarios – variations around default model parameters – were defined to assess the short- and long-term viability of ERS under possible future conditions (Table 1).

Each parameter scenario was evaluated with ERS networks spanning up to 0, 2000 and 6000 km of bidirectional roads, with 40% coverage, with and without policy that effectively enforces ERS use by BETs driving on ERS (labeled “forced use”). The effects of ERS-incentivizing policy were simulated by removing those BET charging strategy options that ignore ERS even when present (see Fig. 1 and Methods).

It is assumed that the upper limit to the build-out rate for static charging infrastructure is unaffected by ERS buildout. All scenarios therefore share the following configuration, unless otherwise specified. Charging infrastructure construction takes place at depots 2025–2045 (capped at 90% of possible locations), rest stops 2025–2045 (100%) and destinations 2030–2045 (30%). ERS coverage is 40% and construction begins in 2030, expanding at up to 1500 bidirectional km (skipping gaps) per five-year period. We believe these time build-out rates are achievable, but they are much more ambitious than current EU HDV GHG emission reduction targets. Destination charging is capped at 30%, as we do not believe it is realistic to assume charging can be offered wherever trucks load and unload.

Charging is capped at 100 kW per vehicle at depots, 500 kW at destinations, 750 kW at rest stops and 375 kW (mean 150 kW including gaps) on ERS, but may be further capped by vehicle battery capacity. BET traction battery usable capacity options are 75₁₆, 150_16,24, 250, 450, 700_24,40,60 and 1000_40,60 kWh, with subscripts indicating capacities only available for a subset of truck tonnages. Total supply of renewable drop-in fuels (biofuels and e-fuels) is capped at 5% of 2020 total fuel consumption, increasing 2% (of volume) annually.

The simulation is computed for consecutive five-year periods, from 2020–2024 to 2050–2054. No electrification takes place 2020–2024 – the period is included to enable comparisons against the present state. Periods are referred to using the first year. The 35 years' time horizon was selected based on typical economic lifetimes of major infrastructure investments.

Table 1

Global model parameter changes in the five parameter scenarios being evaluated. These scenarios are the basis for Figs. 2 and 5. Parameter selection was done based on dominant system dynamics identified in an earlier technical report¹⁹. Other realistic parameter changes are expected to result in less substantial deviations from Neutral. It is beyond the scope of this study to explore what would make these scenarios come true.
Scenario	Parameter changes vs. Neutral
Dynamic-favorable	Vehicle daily time in transit and in use increases annually by 2% (from 9–11 and 12–15 hours in 2020). Daily time at depot and annual mileage are adjusted accordingly. These are all global model parameters describing vehicle time when not in transit. The vehicle population effectively shrinks, as total transport work is unaffected. The utilization rate of depot chargers decreases 1% annually and increases by 1% for the remaining types.
Neutral	All parameters have their default values.
Static-favorable	Triple day-time electricity cost and double ERS construction cost. Half the rate of traffic volume increase (1% from 2%). Utilization of charging infrastructure at depots and rest stops are increased (from 43–63%) to simulate the effects of co-locating the two.
Static-then-ERS	All static charging infrastructure is deployed 2025–2030 and ERS from 2035.
Triple-renewables	Supply of renewable drop-in fuels capped at 15% (from 5%), growing at 2% annually (i.e., triple the annual volume increase).

System Cost and Fossil GHG Emissions

In the simulated scenarios, fossil GHG emission reductions by 2030 (relative to 1990) are 10–32% without ERS and 41–65% with ERS accompanied by policy. The Static-then-ERS scenario enables 51% reduction, with ERS built only later. A 95% GHG reduction by 2050 is only reached in scenarios with forced ERS – which increases electrification – or tripled supply of renewable drop-in fuels.

Figure 2a shows the development of levelized system cost per driven kilometer, including both BET and ICET at shares given by Fig. 2e. The mean cost decreases by 20% from 1.6 €/km in 2020 to approximately 1.25 €/km by 2050. In a business-as-usual scenario with no electrification, the levelized transport cost and total population GHG emissions would increase slightly (Fig. 3a).

Sufficient access to charging infrastructure is a prerequisite for electrification. As scenarios with ERS include more infrastructure than scenarios without, ERS accelerates both electrification and the resulting reductions in mean cost and GHG emissions. The sufficiency criterion is particularly evident in the Dynamic-favorable scenario, where long operating hours and daily distance would necessitate prohibitively large and expensive batteries without access to dynamic charging, making ICETs the cost-minimizing choice on many routes.

If ERS is built and utilization is optional (Fig. 2, black lines), the infrastructure is abandoned in the simulation and becomes a stranded asset beyond 2045 in all scenarios except Dynamic-favorable. Abandonment is the result of a negative feedback loop in the model, where improved BET and battery technology make it slightly cheaper to traverse some routes with trucks equipped with large batteries primarily statically charged at night. When ERS utilization decreases, the cost per remaining user increases, triggering further defection, eventual abandonment and construction of additional infrastructure for static charging. Neglecting to raise user fees would retain users, but not achieve cost coverage. ERS abandonment always leads to higher system cost than if no ERS was built in the first place. In the Static-then-ERS scenario, late ERS introduction causes abandonment of public static charging, but the simulated system cost is largely unaffected as this occurs beyond the outcompeted infrastructure’s economic lifetime.

ERS accompanied by policy (Fig. 2, red lines), reduces mean system cost in all but one parameter scenario, and larger ERS networks bring greater mean reductions than smaller ones. In Static-favorable, the exceptionally large electricity price difference day and night (0.3 €/kWh) result in a mean system cost 1–2% higher in the last periods than if no ERS was built. Such a high electricity cost sustained over decades would be unprecedented in Europe.

The general reduction in mean cost is due to increased electrification, which substantially reduces cost on routes that are not electrified without ERS. Simulated forced-ERS impact on routes electrified also without ERS is within +/- 3% in the Neutral scenario. We have not found any clear patterns in which routes are affected positively or negatively and it is unclear if the simulation is accurate enough for such analysis. Using socio-economic terminology, we conclude forced ERS use is a Kaldor-Hicks efficient policy²⁸ – a relaxation of Pareto optimality – with near guaranteed positive outcome in total and minor negative outcomes for a few agents.

Charging Infrastructure Return on Investment

There are diminishing returns from investments in additional charging infrastructure, and not all types of charging infrastructure yield the same return on investment (ROI – see definition in figure legend). Figure 4 shows the marginal reduction in system cost per marginal increase in total infrastructure investment, given increasing quantities of each type of charging infrastructure, at each stage of the system transition. Depot charging bring the greatest ROI throughout most of the transition, closely followed by a combination of moderate build-out of ERS and public rest stop charging. A combined build-out grants greater ROI than a focused build-out of either. The marginal value of building additional depot and ERS charging is negatively correlated with rate of transport electrification, while public static charging ROI appears insensitive to the stage of transition. High rates of electrification are only achieved in scenarios that include dense deployment of depot charging, combined with some solution for public charging. ERS ROI correlates positively with traffic density.

Monetary savings and GHG reductions are mainly the result of accelerated transport electrification. The marginal value of electrification depends on the levelized cost difference between BEV and ICEV – especially sensitive to electricity and fuel prices, electrical grid tariffs, electricity GHG intensity, GHG valuation, battery and vehicle lifetimes, and short-term battery prices. Future fuel prices are uncertain and influence electricity prices. Electricity prices, GHG intensity, GHG valuation and taxation can differ substantially between EU countries. ERS ROI relative to static charging seems particularly sensitive to daytime electricity prices (lower favors ERS) and battery lifetime (shorter favors ERS). National or regional estimates of the socio-economic benefits will therefore deviate from estimated EU averages in Fig. 2.

Vehicles and Energy Carriers

BETs achieve levelized cost advantage over ICETs once charging infrastructure is sufficiently developed. The simulation does not constrain vehicle replacement rates. At most around 50% of driven distance transitions to BETs during one five-year period in all scenarios. This approximately matches the replacement rate of vehicles resulting from an assumed ten-year economic lifespan. However, to match the timeline of the scenarios, nearly 100% of European new heavy-duty vehicle sales would need to be electric by the second half of the 2020s, or substantial retrofitting would need to take place.

Battery Demand

An unhindered transition also requires that battery manufacturing capacity meets demand, determined by BET adoption, battery pack size, and lifespan. Rapid BET uptake is observed in scenarios with ERS and accelerated static charging infrastructure (Fig. 2). Dynamic charging makes it cost minimizing to reduce traction battery capacities on ERS routes by approximately 65% (Fig. 5). This result extends prior findings that such a reduction is possible for both trucks and passenger cars^{7,10,11,13,14}.

To estimate ERS impact on total battery demand, it is necessary to include all traffic on the entire road network. Charging infrastructure impact on battery ageing should also be considered, as lithium-ion batteries degrade both due to cycling and elapsed calendar time. A capacity reduction decreases calendar ageing cost and increases cycle ageing rate. Dynamic charging in turn partly bypasses the battery to reduce overall cycling. The simulation model doesn't track a vehicle population, making annual new battery capacity estimates challenging, but it allows calculation of the levelized annual replacement rate of batteries in BETs.

ICE Fuel Demand

Sustainable drop-in fuels, including biofuels and e-fuels, remain important. Alone, they offer neither a scalable nor cost effective pathway to full decarbonization, but reaching near zero fossil GHG emissions is not viable in the simulation without them. A complete fossil fuel ban seems attainable by the early-2030s in the Triple-Renewables scenario (Fig. 2C6) and by the late 2040s in scenarios with expansive forced ERS.

For 5% of modelled transport, electrification is never achieved in any scenario. Further research is needed to assess if this outcome is accurate or an artefact of the model or input data.

Charging Infrastructure Demand

Figure 2E-J illustrate how much energy is transferred via each charging method. GHG emissions reductions in 2030 are similar (and in line with targets) in the Neutral and Static-then-ERS scenarios. Between the two, absence of ERS is compensated for by increases in static charging capacity of approximately 220–250% at depots, 130–170% at destinations and 100–220% at rest stops, in approximately twice as many locations.

Without ERS, depot, rest stop, and destination charging contribute around 60%, 30%, and 10% of total energy, respectively. Increasing vehicle utilization makes depot charging gradually less important in the Dynamic-favorable scenario.

In optional ERS utilization scenarios, ERS appears likely to become a stranded asset post-2040. Larger networks are more sensitive to abandonment, even under less challenging scenarios than those presented. The market for static charging is highly volatile in these scenarios. Conversely, in forced ERS scenarios, dynamic charging maintains a stable 40% (small network) or 65% (large network) share of total electrical energy. Policy enforcement stabilizes demand between all charging methods but late ERS rollout adversely affects earlier built public static charging.

Substitution effects and synergies are further explained in Fig. 7.

The primary implication of this study is that ERS – accompanied by policy that guarantees sufficient utilization of the infrastructure – offers a no-regret pathway to greatly accelerated decarbonization of heavy-duty road transport in Europe. This assumes one of the competing technologies emerges as an international standard. Utilization-boosting policy is not needed if technical and economic conditions develop in ways that favor ERS, but policy secures the investment and ensures net benefit effects in adverse scenarios. Benefits include increased BET uptake, reduced mean transport cost, accelerated GHG reductions, reduced long-term battery demand, and a possible ban on fossil fuel use by the late 2040s.

Simulations indicate that a rapid combined buildout of static and dynamic charging can bring heavy-duty road transport in line with EU economy wide GHG emissions targets by the early 2030s, rather than beyond 2050 as currently targeted. ERS would in this case deliver 40–65% of total electrical energy to HDVs. Without ERS, a similar reduction in GHG emissions would necessitate either a substantial and rapid increase in production capacity for sustainable combustion engine fuels, or two to three times as much static charging infrastructure. Neither alternative is preferrable in terms of cost or pollution.

Night-time low-power charging and ERS are identified to be the most cost-effective charging infrastructure types while less than 50% of all heavy traffic is electrified. The results confirm prior findings that battery capacity per BET will decrease by approximately 65% on ERS-equipped routes. We further estimate that ERS contributes a reduction in long-term annual BET battery demand by up to 50%.

Figure Calculations

This section details the steps taken to calculate the data for figures in the main text and elaborates on the reasoning behind selecting what information to show.

Figure 2: Forecasts under five main scenarios

The figure presents forecasts of several emergent properties in the simulated system that are sensitive to changes in input parameters. In addition to the variations made in the included scenarios, formal and informal sensitivity analysis has been performed on several other parameters of potential interest. Parameters that have been analyzed and identified to not influence whether policy can make ERS a no-regret investment include: GHG taxation rate, interest rates, long-term battery prices, and ERS gap ratio and power per vehicle (when above 150 kW including gaps).

In a scenario with triple heavy traffic (closer to central EU conditions), ERS without policy remains in use in the simulation beyond 2050 with the smaller network, but not with the larger network. As seen in Figure 3, charging infrastructure (dynamic and static) constitutes only a small share of total levelized costs for BET traffic and spreading the ERS cost on more users will have little impact as long as the userbase is sufficiently large. Policy to incentivize use is required when charging via ERS is not the option with lowest cost for a sufficiently large population of users and this population naturally increases with increased heavy traffic.

Similar effects would result from inclusion of light traffic, which in total has approximately the same energy consumption as heavy traffic. A less explored effect is that doubling the user base (measured in energy) would make it possible to expand the network geographically without raising the cost per user, as a greater share of the road network would have sufficient traffic for ERS. Expansion of the ERS network should increase its benefits to the heavy traffic. Accurate analysis of the interaction effects between light and heavy traffic would require better access to movement patterns for passenger cars and light-duty vehicles, which has not been within scope of the project.

Figure 3: Distribution of system costs

The figure uses data from two simulated scenarios, different from the set of main scenarios in Table 1, defined to reliably estimate mean transport cost. Mean transport cost per powertrain is challenging to compute in the main scenarios, as the probability that a route is electrified is not independent of the estimated levelized cost on the route.

The no charging infrastructure scenario (“ICET”) in Figure 3 assumes diesel use by vehicles on all routes all years. In the “BET (best case)” scenario, static charging is offered at all depots, all destinations and all rest stops, together with a 6000 km ERS network with 40% coverage, with forced ERS use. When static charging options are unlimited, exclusion of ERS alters the distance-levelized cost by at most a few percent – except before 2030, when no ERS is assumed to be available in any of the main scenarios. Other scenario parameters were the same as for figure 2. Means were calculated as total annual cost across all routes, for each cost component, divided by total annual distance driven by BETs. This method does not capture that electrification also changes the cargo carrying capacity of the vehicle, which in well-optimized scenarios results in a small reduction in the total number of trips required to transport the same cargo.

Figure 4: Charging infrastructure ROI at different fleet electrification rates

The figure uses data from an “infrastructure combination matrix”, different from the set of main scenarios in Table 1. The infrastructure combination matrix is defined to enable calculation of the marginal impact of changes in the composition and availability of charging infrastructure in the transport system. It is a factorial experiment design with 513 scenarios – all combinations of three availability levels for each type of static charging (0%, 25% and 75% of possible locations), three ERS network sizes (0 km, 2000 km, 6000 km), three ERS coverage ratios (25%, 50%, 100%) and three ERS power levels (up to 100 kW, 300 kW, 700 kW per vehicle). ERS use was forced in scenarios and on routes with ERS. Only the 2035 model period was evaluated. The simulation results from the same infrastructure combination matrix constitutes input data to Figures 4, 6 and 7.

For Figure 4, the infrastructure combination matrix was filtered to only retain scenarios with either 0 km ERS, 300 kW ERS with 50% coverage, or 700 kW ERS with 25% coverage. Following filtration, scenarios were paired up such that they differ only in one availability step (none, low, high) for one of the four charging infrastructure types. This left 54 pairs without ERS, 162 pairs with 300 kW ERS and 162 pairs with 700 kW ERS.

Return on investment (ROI), or the marginal benefit from a marginal increase in investment, was calculated by dividing the difference in annualized system cost by the difference in annualized cost of the increased type of charging infrastructure. Because of the optimization logic, an increased offer of a type of infrastructure only leads to an increase in construction if this contributes to reductions in levelized transport cost along at least some routes. Investments become small if additions rarely contribute to cost reductions.

Scenarios were calculated independently of each other, and the ROI calculation treats substituted infrastructure as if it was never built. The value of stranded assets should be accounted for if the substitution takes place within the expected economic lifetime of the assets. Doing so could reduce the system-level ROI below zero for some investments. The figure also does not consider that different stages of the transition in practice correspond to different years with different cost and technology conditions, or that conditions (such as battery performance) change over the infrastructure lifetime.

Figure 7: Competition effects between charging infrastructure placements

The infrastructure combination matrix from Figure 4 was reused to estimate the marginal change in total electrical energy delivered via one type of charging infrastructure resulting from a marginal increase in the availability of another type of charging infrastructure, all else kept equal. The dataset allows pairwise comparisons in two availability steps – from no to limited availability (e.g., depot charging offered at 0% or 25% of route origins) or limited to high availability (e.g., depot charging offered at 25% or 75% of route origins).

The purpose of the figure is to explain the risk that investors take when the future availability of competing infrastructure is uncertain. It is clear that the public sector should strive for maximum transparency regarding its planning and decision-making processes related to ERS buildout. Without transparency, private sector investors in public fast charging infrastructure must compensate for the increased business risk through raised prices, which reduces the financial incentives for vehicle owners to make the transition from ICET to BET.

MOSTACHI

Summary

MOSTACHI (Model for Optimization and Simulation of Traffic And CHarging Infrastructure) is a free and open-source software tool designed to quantify effects that emerge from dynamics in a complex system of road-bound traffic on a national-sized road network or larger, during a gradual transition from internal combustion engine vehicles (ICEVs) to battery-electric vehicles (BEVs). The tool is uniquely capable of modelling interaction effects in time and space between:

Dynamically placed competing installations of several types of static and dynamic charging infrastructure, priced and sized for balanced supply and demand,
Multiple vehicle classes operating on hundreds of thousands of overlapping transport routes,
Driveline (ICEV or BEV), charging behavior and battery capacity, jointly optimized to minimize total internalized cost per transport route.

The simulation tool is designed to handle high spatial resolution (based on a segmentation of the real road network) and mixed temporal resolution (modifications of vehicles and infrastructure in five-year periods, and simulation of energy consumption, charging, battery state and driver rest on a per-road-segment basis). MOSTACHI is intended to be used to study how the transport system will develop under different future external conditions. Following simulation of multiple scenarios, comparative analysis is possible to understand how changes in input parameters affect emergent properties in the complex system. Such input parameters can be energy and fuel price trends, battery technology development, interest rates, changes in vehicle utilization rate, infrastructure subsidies, electrical grid expansion rate, or course-altering decisions such as national build-out of electric road systems (ERS).

Running the tool only requires access to a type of national transport flow dataset that is available publicly or on request for many countries within Europe and internationally. This enables reuse of the tool to replicate and alter the present analysis in other geographical contexts. It takes a few minutes on a moderately powerful 2020 laptop to compute one scenario, developing over seven five-year periods and including four vehicle classes operating on a few hundred thousand transport routes on the Swedish national road network (approx. 200,000 road km).

Model Design and Architecture

MOSTACHI’s simulation logic is conceptually very simple. Initially, it requires defining a) global model parameters that represent the national context, such as expected electricity prices and interest rates for the studied period, and b) one or multiple scenarios that specify the upper limit of the scope of charging infrastructure expansion. Subsequently, the model iterates over the desired timeframe (currently 2020-2050 in 5-year increments) and for each period it performs independent cost-minimization for overlapping transport routes. For each route, it is decided whether a diesel or battery electric powertrain would offer the lowest total transport cost, and for electric powertrains, the optimum combination of battery size and charging locations. The model then sums the demand for charging (peak electrical power) at all locations considered in that time period, adjusts the estimated pricing of charging at each location, and repeats the calculation with the new price information. When charging patterns have stabilized, charging infrastructure is fixed for that period. Built infrastructure is inherited from one period to the next, but vehicles are not, as it is assumed that vehicles can be re-sold to operators in other locations. Charging infrastructure is permitted to be downsized or decommissioned beyond the economic lifetime used to estimate user charges.

MOSTACHI models four categories of charging infrastructure for electric vehicles: dynamic charging (electric roads), and three forms of static charging (truck depot charging, route destination charging, and charging at major driver rest stops). The model accounts for economies of scale for all types of charging infrastructure, i.e., that larger facilities yield a lower cost per energy unit given the same rate of utilization. The extent of such scale benefits varies, being minimal for depot and destination charging, modest for rest stop charging, and substantial for ERS.

Simulation of one five-year period takes place as follows. 1) Any already built charging infrastructure is inherited from the previous simulation period. 2) According to limits imposed by the scenario specification, charging infrastructure at all permitted locations is treated as if it was built. 3) The price of charging at each location is estimated using a heuristic. 4) For each transport route and vehicle class, the route is traversed one road segment at a time to identify which combination of driveline (ICEV or BEV), battery storage capacity and charging strategy would result in the lowest total annual cost of transporting all goods along that route. 5) All charging for each charging infrastructure location is summed, resulting in an aggregate demand, which can be above or below any inherited capacity. 6) The cost of charging at each location is updated based on estimated utilization and any inherited overcapacity. Once past the depreciation period, installed capacity can be reduced to match demand. 7) If the cost of charging at a site is found to be more than five times the global average for that charging infrastructure type, that site is removed as a candidate of infrastructure construction in this period. 8) The simulation is repeated, until the total energy delivered per type of charging infrastructure has stabilized. The simulation flow is represented visually in Figure 2.

The iterative optimization until convergence within time periods, of vehicle configurations, charging strategies and charging infrastructure expansion, is a computational way of handling that a) the cost of charging at a location depends on how much charging takes place there, b) where charging takes place and how much energy is purchased during a charge session depends on the vehicle battery’s total storage capacity and current state of charge, and c) both optimal battery capacity and current state of charge depend on available charging infrastructure and pricing per location.

MOSTACHI embodies a free-market model, representing independent decision-making processes by vehicle owners regarding vehicle purchases and charging locations. It also allows for competing charging infrastructure established in different locations by multiple operators. While more resource-efficient system configurations might exist than those resulting from this simulation, it is unclear what incentives could steer an uncoordinated market economy towards such a global long-term optimum.

Model Capabilities and Limitations

The following real world system dynamics are captured in MOSTACHI:

Total amount of traffic and energy consumption per vehicle change for each time period.
Traffic is only electrified when and where switching to an electric vehicle reduces cost relative to operating a combustion engine vehicle in the same year. All vehicle operators strive to minimize their own costs.
The point of break-even between BET and ICEV operation is an emergent property of economic parameters, state of technological development, operating patterns, charging infrastructure along the route and possibilities to share the cost of that infrastructure with other routes.
Vehicles traversing different routes share public infrastructure. Routes sharing an origin or destination share depot and destination charging infrastructure.
Charging stops along the route raise transport cost if they do not align well with mandated driver rest times.
Vehicles of different weight classes have different energy consumption and gain different range from the same charging power.
Not all battery capacity options are available for all vehicle weight classes.
Battery power scales with battery capacity and vehicle battery capacity cannot be reduced below minimum power output requirements.
At each simulation step along a route, charging is capped by current state of charge (limits energy), vehicle battery size (limits power) and charging infrastructure power (limits power).
Raising the maximum power of charging infrastructure makes it possible to deliver more energy during the same charging session length, but also increases battery wear. Maximum charging power is also limited by the battery capacity, with larger batteries able to receive more energy per unit of time.
Very high ERS power quickly leads to full batteries, reducing power demand on later ERS sections.
Energy charged on ERS extends battery range outside the ERS network.
Routes passing orthogonally to an ERS stretch do not benefit at all from ERS.
Charging and discharging c-rates (charging power relative to battery capacity) affects battery lifetime.
All battery cycling contributes to battery ageing. As ERS can bypass the battery to deliver energy directly for propulsion, it reduces cycling.
Battery capacity, type and location of charging are selected jointly to minimize cost, per route.
Rapid advancements in battery technology and the resulting plummeting of levelized battery costs have effects throughout the system, most prominently reducing the marginal value of infrastructure that enables battery capacity reductions.
Installed battery capacity affects cargo carrying capacity, which affects the number of trips required to transport cargo, which affects driver and vehicle costs.
Only charging infrastructure that has a sufficient customer base to be profitable is built.
New charging infrastructure can both complement and compete with other charging infrastructure along all intersecting routes, with ripple effects.
Charging infrastructure of all types takes time to build out and options become denser over time.
Charging infrastructure that becomes available later can outcompete charging infrastructure that was built earlier.
Changing economic conditions or technical properties can lead to changes in behavior throughout the lifetime of an infrastructure installation.
ERS infrastructure must achieve critical mass to be cost competitive and to be built but may lose that critical mass in later periods, raising costs for remaining users.
An early national decision to build ERS comes with added economic risk (due to current lack of standardization), but a late decision risks turning already built static charging infrastructure into stranded assets.
Expanding a small ERS network increases utilization of already built ERS segments, because ERS-adaptation of vehicles becomes a cost saving decision along additional routes. However, there is a threshold when further ERS expansion reduces utilization, because the marginal benefit becomes smaller than the marginal cost. This threshold depends on a range of factors, including ERS cost, traffic density, traffic patterns, costs and availability of competing charging infrastructure and battery costs.
Usage fees are updated for each five-year period, to cover the annual levelized infrastructure cost at current utilization rates. ERS is priced globally (same price everywhere) and static charging is priced per site. See model calculations.
The mixing ratio of fossil and renewable diesel is dynamically calculated based on total consumption and a predefined supply of renewable fuel.

The following known real world system dynamics are not captured in MOSTACHI:

Availability of charging infrastructure outside of the simulated geographic region (Sweden, in this study) will influence demand for charging inside the simulated region. The current handling of this is that parts of routes that fall outside of the region are assumed to have access to equivalent charging infrastructure as the part inside the region.
Vehicles manufactured in different years with different technologies operate on the roads simultaneously. In the model, all vehicles are replaced every five-year simulation period, purely to keep the computational complexity manageable.
There may be upper limits to how quickly the total population of vehicles can be replaced. Historical rates of vehicle replacement may not be a good predictor of this rate, as they have been driven by tightened emissions regulations in Europe. MOSTACHI only handles vehicle replacement rate limits indirectly through limitations in charging infrastructure build-out rates in scenario definitions.
Vehicle and battery lifespans may in reality be coupled, if batteries are integrated as a structural component. The two are decoupled in the simulation, based on reasoning that if battery and vehicle lifetime differ substantially, batteries must become an easily replaceable (or reusable) spare part.
The price elasticity of transport demand implies that changes in cost will alter traffic density throughout the road network. This effect could be substantial and may affect charging infrastructure economics, travel time, route selection and may result in changes to the road network itself.
Route selection may be influenced by charging infrastructure availability. Routes are fixed in the simulation due to computational cost constraints. We find it very unlikely that so much traffic would be diverted from minor roads to major roads (where most charging infrastructure is placed) that this would have a substantial impact on the economics of the charging infrastructure.
Vehicles that must operate on multiple routes need to be adapted to the joint constraints imposed by all these routes. Optimization is performed independently per route, though dependencies arise through joint infrastructure utilization. See discussion on input data below.
Both temporally varying electricity costs and congestion at charging infrastructure will create economic incentives to operate during other times of day than today. Such adaptation will contribute to increased utilization rates and reduced cost of charging.
Aggregate battery demand will affect battery prices if the geographic system boundary is assumed to be sufficiently wide. The same may be true for fossil and renewable fuels and electricity.
Access to dynamic charging creates a stronger economic incentive for reducing the total time a vehicle stands still, while high dependence on static charging creates a requirement for vehicles to remain unused for a portion of the day, especially if low-power night charging is substantially cheaper than high-power daytime charging. Vehicle utilization is not a dynamic model parameter.
Future availability of autonomous trucks may alter the demand for different types and placements of charging infrastructure, and the availability of different charging infrastructure may influence how easy it is to introduce autonomous trucks on the market.

Terms and Definitions

Here follows a list of explanations of terms used in the methods section and main text. Definitions of costs and model inputs are found in separate sections.

Simulation model, input data, scenario, experiment: The logic represented by the MOSTACHI source code is what we refer to as the simulation model. Input data used by the model are numeric constants (parameters) and transport routes, and pre-processed intermediary representations of these data. The input data represent the world to be analyzed. A MOSTACHI scenario is a combination of

maximum charging infrastructure availability per simulation period (including size of the ERS network),
ERS network electrification rate (gap ratio),
maximum charging power per vehicle per type of infrastructure,
available battery capacity options per vehicle type,
available charging strategies,
deviations from the default parameter values.

Experiments are conducted by computing several scenarios and comparing differences in simulation results.

Road segment, depot, destination, rest stop, charging station: A road segment is the graph theoretic representation of a short section of road (edges) between two road network intersections (nodes). Truck depots, destinations and rest stops are treated as road segments without length. Depots (route origins) and destinations exist in a fine-mesh grid, independent from the road network graph, and each grid cell is treated as one charging infrastructure site in the cost model. The exact coordinates of rest stop parking areas are used to link these to all road segments within one kilometer (0.62 miles) distance. The term charging station may be used interchangeably with charging infrastructure placed at rest stops.

Charging infrastructure (static, dynamic, offered, built): Divided into infrastructure for charging vehicles while stationary (static charging infrastructure) and in motion (dynamic charging infrastructure). Includes costs for hardware, installation, maintenance, establishing a connection to the electrical grid, operational grid charges, and capital interest. All costs are summed over the infrastructure lifetime (discounting period) and annualized over the parameterized infrastructure lifetime. User prices are determined assuming the within-period utilization will be sustained throughout the lifetime. This can lead to underpriced utilization early in the lifetime, if other competing charging infrastructure is built later that makes the economic lifetime of the site much shorter than its expected discounting period. Such inaccuracies may affect charging behavior in the simulation but have less impact on estimations of total system cost. The model distinguishes between offered infrastructure and built infrastructure, where an offer means that the scenario definition permits infrastructure to be built in a location if it is in demand, and built infrastructure is the simulated outcome of a period that contributes to system cost.

ERS network, electrification rate: The electric road network is the subset of the entire road network where dynamic charging via ERS is supported. A scenario parameter for the ERS network is its electrification rate. An electrification rate of 40% means that physical ERS infrastructure is built on 40% of the total lane distance of the ERS network, interleaved by non-electrified gaps. If the power per vehicle is doubled and the electrification rate is halved, the effective rate of energy transfer per vehicle remains unchanged. ERS network sizes refer to bidirectional road distance including gaps. Reducing the electrification rate reduces the cost of the ERS infrastructure but increases costs in other parts of the system.

Route: An unbroken one-way sequence of road segments describing a directed movement along the road network, from origin to destination. Each route is associated with metadata that describes the annual number of tons of goods transported and number of truck transports, per vehicle weight class. Effective load factors vary and as far as we can tell from the data provider, number of transports includes trips without cargo. The first road segment in a route is always a depot and the last is always a destination. The first time a route passes near a rest stop, the zero-distance rest stop is inserted into the route (but the truck may not stop there).

Vehicles, trucks: In the present study, MOSTACHI has been applied to simulate freight traffic with medium and heavy-duty trucks. However, the traffic data used contains no explicit information about individual trucks. Trucks only exist in the model as virtual agents that start at a depot, traverse a route (possibly including rest stops) and stop at a destination. Vehicles are always assumed to be equipped with the battery and vehicle technology of the current five-year period for which the simulation is calculated: a departure from reality made to simplify the calculations. Vehicles that utilize ERS must be equipped with a separate ERS pickup, which adds cost.

MGV16, MGV24, HGV40, HGV60: Trucks in four weight classes are included, using definitions inherited from the input transport data. MGV16 is a medium-heavy two-axle goods vehicle (truck without trailer), total weight 3.5-16 tons, which is usually used in local distribution. MGV24 is a medium-heavy three-axle truck without a trailer, total weight 16–24 tons, which is usually used for construction work. HGV40 is a heavy two-axle tractor and three-axle trailer, total weight 25–40 tons, which is usually used for long-distance transport. HGV60 is a heavy three-axle truck with a four–axle trailer, total weight 40–60 tons, which is used, for example, for transporting round timber²⁹.

Battery pack: Battery electric vehicles are equipped with battery packs, which are modelled separately from the rest of the vehicle, both parameter- and simulation-wise. The usable (net) storage capacity (in kWh) of vehicle battery packs is determined per route, vehicle class and time period, based on a) the maximum output power per battery capacity for that simulation period, b) the minimum required output power for that vehicle weight class, and c) the capacity that minimizes the total cost of transporting the annual cargo. Conversions between usable (net) and installed (gross) storage capacity are made via the state of charge (SoC) window parameter, which represents the fact that current battery chemistries perform so poorly near minimum and maximum SoC that these are artificially hidden by the manufacturer from vehicle users. Technical sources of values for battery parameters, such as weight per capacity, usually state gross figures.

State of charge (SoC): The energy level in a battery pack is referred to as state of charge, reported in percentage of net capacity or kWh. The SoC is updated sequentially for each road segment along a route. During dynamic charging, the SoC is updated based on the difference between supplied power and propulsion power. The SoC is initiated at 50% at the depot, followed by an opportunity for depot charging limited by the route's share of total daily operating time. Depot charging may not be available due to lack of infrastructure or an incompatible charging strategy. If the energy level anywhere along the route reaches a minimum threshold for remaining range, the current combination of battery capacity and charging strategy is rejected.

Charging strategy: Along a route, charging infrastructure can be offered at many locations, resulting in different direct and indirect costs. Successful traversal of a route may require utilization of all or a subset of offered charging locations, with outcomes dependent on battery capacity. The charging strategy defines which locations are considered:

NA_Diesel: The vehicle is equipped with a diesel powertrain – charging infrastructure is ignored.
Depot: The vehicle only charges at the depot.
DepotAndErsCharging: The vehicle charges at the depot and ERS. Becomes ErsCharging if depot charging is not available on the route.
ErsCharging: The vehicle charges only from ERS infrastructure.
AllPlannedStops: The vehicle charges at the depot, destination and when the driver takes a legally mandated break after at most 4.5 hours of driving. Breaks are taken up to 60 minutes earlier if charging is possible at a rest stop. Becomes PublicStaticCharging if depot charging is not available on the route.
PublicStaticCharging: The vehicle charges only at rest stops and destinations. This strategy is not tested separately.
AllPlannedStopsAndErs: As AllPlannedStops, but the vehicle can also charge from ERS.

A vehicle cannot charge at a type of charging infrastructure if no such infrastructure is offered along the route, regardless of charging strategy. With all strategies, driver breaks are taken up to 30 minutes earlier if a rest stop is available, else a road-side stop takes place. This logic represents that the majority of real stops for driver rest do take place at designated rest areas (conclusion from private communication with Patrick Plötz, who had access to stop data from the major European truck OEMs for use in a study mapping and characterizing stop locations³⁰).

Charging power and C-rate: When a vehicle is charging while stationary, this takes place at the lowest of the infrastructure's maximum power per vehicle, the battery's maximum allowed charging power and the lowest power required to reach full charge before departure. When a vehicle charges dynamically from ERS, this rate is capped by the per-vehicle maximum power of the infrastructure minus power for propulsion, and further capped by what the battery can receive. The maximum charging rate of a battery is a parameterized (period-dependent) multiple of the effective C-rate, with C-rate being (input or output) power divided by storage capacity. This means only vehicles equipped with large battery packs may be able to fully utilize high-powered charging infrastructure, when such is available.

Model Calculations

This section describes how different cost components are calculated and how they are related.

System cost: The sum of all levelized costs of vehicles, batteries, charging infrastructure, capital interest, fuel/electricity, driver salaries and GHG emissions (internalized and externalized, or taxed and untaxed). Excludes road and pollution taxes. All costs are in 2020 Euro; no conversions are made to net present value or to adjust for inflation. Interest rates are considered as rates above inflation.

Transport cost: The sum of all levelized costs affecting the vehicle operator – vehicles, batteries, capital interest, fuel/electricity and distribution, driver salaries and all taxes. Residual value is subtracted for vehicles and batteries.

Weight adjustments: Electric vehicles will differ in weight compared to the original diesel vehicles. Cargo carrying capacity adjustments are made to the per-route traversal costs to compensate for that a different number of annual trips are required to transport the same total amount of cargo with maintained fill-rate. Loading capacity is often limited by cargo volume rather than weight – therefore cargo capacity is modelled to change by half the weight difference between ICEV and BEV. Fuel (energy) consumption is not adjusted for weight differences.

Vehicle cost: The sum of purchasing price and maintenance costs minus residual value, levelized over the assumed economic lifetime. Vehicle cost excludes battery costs.

Battery cost: The purchasing price minus residual value, levelized over the simulated lifetime. Purchase price depends on installed usable battery capacity and net battery price in the current simulation period. Net battery price is gross price (what is normally listed in sources) divided by SoC window size.

Residual value at depends on battery state of health (SoH) and year of decommissioning. Batteries degrade from passed calendar time and use, beginning at 100% SoH and ending at 0% SoH when the battery can no longer store any meaningful amount of energy. The market value of used BET traction batteries, at a reference SoH (80% in this study), has been approximated using an S-curve, from zero value in 2020 to 30% of the price of new batteries in 2050. This emulates a maturing second life market and recycling industry. Batteries remain in vehicles until any of four conditions is met – the vehicle reaches its end of life, 80% SoH has been reached, remaining battery capacity is insufficient to complete the route, or the maximum power output is insufficient to meet the requirements of the truck. The most common condition is in later years is end of truck lifetime. Decommissioning before 80% SoH reduces the value loss proportionally to the improved health.

Battery ageing is approximated using the sum of calendar ageing – at a constant rate per calendar time – and cycle ageing. Cycle ageing takes place when energy flows into and out of the battery, calculated for each traversed road segment along a route as where a_cyc is the percentage of capacity loss from cycle ageing, e is energy flow (in kWh), c is c-rate, and c_ref is reference c-rate. C-rate is power divided by battery capacity, such that a c-rate of one means the battery is charged or discharged in one hour. On a single simulated road segment (which includes stops), either charging or discharging takes place – batteries are conceptually like water tanks and energy can only flow in one direction at a time. Charging and discharging are capped at multiples of c_ref. The constant 1⁄2 is included because a full cycle consists of both charging and discharging and ¾ is a lower bound for the rate of cycle ageing. When total ageing sums to one, the battery has reached the reference SoH and is decommissioned.

This simple equation has no direct source and approximates what the main author has understood about battery ageing mechanisms from battery experts and literature (see references in the supplemental parameter tables). The equation has been designed to a) use variables available in the model, b) allow parameterization of the aging rate, c) let higher charging power result in more rapid battery ageing, and d) let very low charging power still result in some ageing. No attempt has been made to capture ageing effects that are not primarily controlled by changes in charging infrastructure design, such as battery temperature. Capturing the influence of current battery state of charge (SoC) on battery ageing would likely improve results, but would necessitate much more complex charging strategies.

The battery aging model is a potential weak point in the study, as modeling of battery ageing is still an active area of research. However, since most batteries are decommissioned at the end of the vehicle lifetime, flaws of the battery ageing model should only impact the main conclusions if battery ageing is grossly underestimated.

Charging infrastructure cost: A parameterized linear model is used to calculate infrastructure cost, with a fixed cost per site and an increasing cost per kW of installed peak charging capacity (site total). Maintenance costs are assumed to be proportional to installation costs, applied yearly and profit margins are applied. The initial cost of connecting to the power grid is included, as are operational grid fees. All costs are levelized over the discounting period for that type of infrastructure.

ERS installation costs are penalized by a risk multiplier for early simulation periods, to penalize construction before Europe has agreed on a common technical standard. Early construction could result in a need to later re-invest in infrastructure that has already been built. The risk is high in model year 2020 and then decreases rapidly.

Charging infrastructure pricing: The annualized levelized infrastructure cost of the charging infrastructure is spread across the annual sold energy within the simulated time period. Static charging fees are calculated per site and ERS use is subject to a shared fee across the entire ERS network.

Fuel or electricity cost: Fuel costs (parameterized) are pre-calculated per simulation period based on assumed costs of fossil and renewable diesel fuel and the blend ratio between the two. AdBlue costs have not been included. The cost of electricity (parameterized) is specified as a night- and daytime tariff per electrical grid region. Separate utilization curves are used for each charging infrastructure type to calculate the mean electricity price for the entire day. This means that depot charging is closer to night-time prices (assumed lower) and the remaining charging types are closer to day-time prices (assumed higher).

CO₂ emissions cost: Greenhouse gas emissions in the model originate from diesel consumption, electricity consumption and battery production. Diesel emissions depend on the blend ratio of fossil and renewable diesel. Part of the social cost of carbon is internalized through a carbon tax applied to emissions from fuel and electricity. Emissions from battery production are not taxed and any such tax would have a negligible effect on transport economy at current EU CO₂ valuation.

Post-hoc adjustment of renewable fuel ratio and CO₂: The mixing ratio of renewable and fossil diesel is estimated in advance for each simulation period, but re-calculated at the end of the simulation based on an assumption that renewable fuels are maximally used up to a supply cap. This results in adjustments to total costs of fuels and emissions and estimated emissions.

Driver cost: Applied per time unit, excluding time at the depot. We assume one driver per truck.

Capital interest cost: Interest is applied to all capital investments, assuming a flat amortization over the discounting period. Three different rates are applied depending on assumed financial risk, with the lowest rate used for state-backed ERS investments and the highest rate for depot charging infrastructure that is built for one company and cannot easily be moved.

Taxes: Today's taxes on diesel fuel have been separated to the best of our ability into CO₂ tax, pollution tax and road tax, where the road tax is applied equally to diesel and electricity based on the HGV40 energy consumption rate for each simulation period. Road taxes are thus independent of choice of driveline. A separate pollution tax is applied in later simulation periods, on assumption that society will want to penalize the additional particle and noise emissions of ICEVs once BEVs have become ubiquitous.

Model Validation

Numeric variables with strongly typed units (e.g., kWh per kilometer) are used throughout the program source-code to minimize the risk of erroneous calculations. Informal validation of the model has been performed by ensuring that input data passes through without distortion, that population totals approximately match independently calculated estimates for the same values and that values are reasonable in proportion to each other. Visualization techniques have been used extensively throughout the development to identify geographical and temporal anomalies in the output data. To the extent such comparison has been possible, simulation results have not been observed to disagree with findings in prior research by other research groups.

Input Data

Global Parameter Values

A guiding principle for the development of MOSTACHI has been to avoid built-in assumptions about what a future electrified transport system should look like, in particular how electric vehicles of different classes and in different operational patterns should be charged. The model therefore uses a large number of numeric input parameters with separate values for each five-year period. These represent expected battery technology advancements, fuel and electricity prices and emissions, charging infrastructure construction costs, vehicle costs and performance and other properties that are independent of the system dynamics and for which there are generally much more reliable forecasts available than for what we aim to simulate.

Input parameters can be changed as part of defining scenarios to be simulated, which facilitates sensitivity and robustness analysis and exploration of how the system dynamics are influenced by different possible societal trends, such as increasing or decreasing density of traffic or relative pricing of electricity and diesel. Some parameters may require new values if the model is applied to other geographic areas or vehicle segments.

Parameter tables are provided as supplemental material and enclosed with the MOSTACHI source code.

Traffic Patterns

Here follows a summary of the methodology used to prepare traffic pattern data for this study. Except for a few methodological improvements, detailed here, the full methodology, all validation metrics and references to validation datasets are described in a prior technical report²⁶.

In absence of directly measured traffic data, transport route data in the form of an origin-destination matrix was exported using the 2017 base scenario from the Samgods transport simulation²⁷. Samgods is an independently developed and extensively validated Swedish national model of freight transport within, into, out of and through Sweden. Input to the Samgods model is survey data capturing the annual flows of goods using all modes of transport (road, rail, sea and air) between all pairs of Sweden’s 290 municipalities, plus major foreign regions and key transport hubs such as ports. Samgods then distributes these goods flows on modes of transport, including rail, sea and five different truck weight classes, of which we have retained four (MGV16, MGV24, HGV40 and HGV60) due to perceived data quality issues with the last (LGV3). The input dataset encompasses around 200,000 freight flows, further split into yearly tonnage per vehicle weight class. We observe that Samgods generally assigns goods to lighter vehicles on short routes and heavier vehicles on long routes, but refer to its documentation for details.

This base scenario dataset was shared with us by Magnus Johansson at The Swedish National Road and Transport Research Institute (VTI). Data in a similar format capturing transport patterns across all of Europe have since been made available by others³¹, which should make it possible to apply this methodology to directly study effects across all of Europe.

To convert region pairs to routes on the road network, a point coordinate was sampled within each origin or destination region (hubs excluded), independently for each flow. Points were sampled based on openly available polygon data³² representing commercial non-retail zoning areas in Sweden, under the assumption that most truck freight begins or ends within such an area. Polygons were weighted based on provided meta-data about type of industry, number of businesses and number of employees within the areas. A note here is that during validation, a high-resolution raster of population density was used in place of the polygon data to sample origin and destination points. The switch to use commercial zones was motivated by that population density gives sampled points within residential rather than industrial areas within cities, which does not match how heavy-duty trucks are used. This switch should only affect within-city route selection and should have no impact on the results presented in this study.

After enriching the transport flows with start and end coordinates, the Open Source Routing Machine³³ was utilized to determine the shortest route along the OpenStreetMap road network for each coordinate pair. This produced long sequences of very short segments, together representing unbroken routes. The resulting graph made up of all these segments was simplified by merging sets of segments that formed a non-branching chain into one segment, retaining (and combining) meta-data about segment length, speed limits, directionality and lane count. Potentially useful road network properties such as curvature, road gradient, and number of lanes were not available with sufficient quality in OpenStreetMap to be of use.

As the input dataset only contains traffic that involves the Swedish road network, it has not been possible to calculate any charging infrastructure economics outside of Sweden. Routes have been truncated at the border in the simulation and the route distance outside the border has been assumed to have access to charging infrastructure with the same density, ratios and pricing as the part inside Sweden.

Validation and calibration of the resulting route dataset was performed against a) national statistics for total transport assignments, total moved tonnage and total driven distance, b) statistics on truck fill rates, and c) interpolated measurements of real traffic throughout the road network. There are minor but important differences in what traffic and vehicles are included in the Samgods data and in the national reference datasets, which complicates the validation work. Validation against road traffic took place by aggregating the route data into annual vehicle passages per road segment, then by further computing the maximum traffic intensity of any segment intersecting each cell in a fine-mesh spatial grid. The same rasterization was performed for official interpolated measurement data for the entire Swedish road network and the two rasters were visualized as geographic bar charts to study spatial correlation and identify geographic discrepancies.

Two calibration steps were taken: 1) to increase the number of trips made per route to more closely match the real distribution of load factors (which effectively introduces return trips without load), and 2) to jointly rescale number of trips and tonnage per route by common scale factors for all routes in different length buckets, such that the sum of traffic in all buckets more closely matched real measured traffic on the road network. After calibration, the distribution and volume of traffic along the road network closely matched real numbers, except for underestimates of total traffic within major urban centers. We speculate that this is explained either by bus traffic – absent in Samgods but present in measured traffic – or by that Samgods undersamples short-distance transport assignments. We have not come up with any way to confirm or reject these hypotheses. Total travel distance and number of trips were then 40% above national statistics and total tonnage 10% above national statistics, which we believe is approximately correct given that the statistics – but not Samgods and measured traffic – exclude foreign-registered trucks, known to make up a substantial ratio of total traffic but a smaller ratio of total cargo tonnage.

The transport pattern input data lacks information about the temporal distribution of when routes are traversed, and which routes are traversed by the same vehicles and in what order. It is partly because of this limitation that cost optimization takes place per route. Furthermore, this limitation leads to charging infrastructure utilization rates (ratio of peak installed power utilized per hour of the year) being estimated globally per charging infrastructure type and given as an input to the simulation. The same holds true for vehicles – as no information is available to the model about individual vehicles across transport routes, economic calculations in this study cannot account for effects due to variations in vehicle utilization within the vehicle population, from requirements to traverse routes with very differing properties, or operation during unusual times of the day.

While low-bias GPS data from well-sampled large vehicle fleets could potentially yield a more accurate model of today's transport system than the route data employed here, historical data would not account for changes in transport planning (assignment of goods to vehicles) arising from changed conditions along the road network. Furthermore, GPS data is difficult to access for researchers and transport planners worldwide and methods requiring this data may be more challenging to scale. Furthermore, if unbroken GPS traces are accessible, it is straightforward to transform such data into the route format used in MOSTACHI, and the model could with relative ease be adapted to handle multi-stop routes. In short, route-based analysis fails to capture some of the operational characteristics of commercial trucks and truck fleets, but historical GPS data would only partially resolve this.

Order of ERS Deployment

The utility of ERS infrastructure depends both on where on the road network the infrastructure is placed and the size of the ERS network. Maximizing the lifetime utility of an ERS network requires not only that an optimal layout of the network is determined, but also that the step-by-step order of installation is optimal. We do not attempt to solve this very difficult optimization problem, but we compute a build-out order that achieves greater utilization rates than the naïve solution of selecting the roads with greatest average annual daily traffic (AADT). When a MOSTACHI scenario is defined to include an ERS network of size X km, this pre-computed buildout order is used to select on which segments of the road network ERS should be offered. A subset of offered road segments are then selected for ERS construction during the demand-driven placement of infrastructure.

The expansion order used for ERS in the present study has been calculated using Algorithm 1, described in pseudocode. This order maximizes synergies between already built ERS and new ERS but does not take static charging infrastructure into account. The definition of synergies here is that the next road segment to build is that which is most traversed by vehicles using the already built segments. The starting point of the ERS network was set in the south of Sweden, in order to maximize the synergy effects with electric roads in neighboring countries. A map of the build order is available in a prior technical report¹⁹.

Create an index I_A:(S_a,S_b)àW_ab, b>a of association strength W_ab between pairs of road segments (S_a,S_b)∈S

Create an index I_T:R_xàT_x of total number of annual trips T_x for each route R_x∈R

Discard all R_x with trip count T_x < 150 or segment sequence length |S_R| < 100

Build an index I_R:C_aà(C_b,S_x[])[] of all routes R_O that go between two grid cells (C_a,C_b)∈C

For each grid cell C_i, in parallel:

For each route R_j departing from C_i:

Make a set S of all the segments traversed by R_i

If a route R_r pointing in the opposite direction to R_j exists:

Get R_r from I and add the segments in R_r to S, with repetition

Take a random sample S^* of 100 road segments from S

Remove repeating items from S^*

For each pairwise combination (S_a,S_b) of segments in S^*:

Increment the association strength W_x in I_A for (S_a,S_b) by T_x from I_T

After every 10k pairs of (C_a,R_a), pause the parallel loop and:

Discard all (S_a,S_b)∈I_A where W_ab < max(W)/1000

Discard all (S_a,S_b)∈I_A where W_ab < 2000

For each segment S_a in I_A:

Discard all but the top 100 associations (S_a,S_b) from I_A in descending order by W_ab

Build a graph G representing the node-edge relationship of I_A:

Edges E_ab∈E represent a road segment pair (S_a,S_b) with weight W_ab

Set E^a∈E contains all edges connected to node S_a

Set E^m∈E^a is the subset of edges that connect S_a to another already marked segment S_b

Nodes S_a∈S represent road segments, with properties V_a, V_a^*: and marked∈[true, false]

V_a = ∑W_ab over E^a (kept constant)

V_a^* = log(V_a) * ∑W_ab over E^m (updates when a neighbor gets marked)

V_x^* is initialized to a random number r∈[0,1]

Start with a seed of marked road segments

Create a list L of all nodes, sorted in ascending order of V_x^*

Create a list O containing the segment build our order

While |L|>0:

Mark the last item S_x in L and remove it from L

Append S_x to O

Resort L

Algorithm 1: Calculation of the build-out order of electric roads

Rest Stop Locations

Independent prior analysis based on GPS data from connected trucks³⁰ has identified 424 locations within Sweden where many trucks from several manufacturers stop. These locations, primarily rest areas along major motorways, are used as options for placement of public fast charging stations. A map of the locations is available in a prior technical report¹⁹.

No pre-computation of build-out order was performed for static charging infrastructure, which is instead offered in random order (the same order for all experiments). While it is likely that sort orders achieving greater utility can be found, centralized coordination of privately financed investments in charging infrastructure was not deemed a realistic approximation of reality.

Output Data

Simulation of a MOSTACHI scenario results in the generation of three data files containing columnar tab-separated values. Although the simulation runs in five-year periods, output values represent annual data.

A summary of key metrics for the entire simulated system, including total CO₂ emissions, total annual system cost, total annual cost components for ICEV and BEV traffic, share of electrified transport work per vehicle class, total energy delivered per type of charging infrastructure and charging strategy, and total cost per type of charging infrastructure.
Information about the location, type, installed capacity, utilized capacity and cost of use of each charging infrastructure object present in the simulation. Data for ERS infrastructure is reported in a 1 km² raster, while static charging infrastructure has an exact coordinate per site.
Total annual traffic (vehicle passages and tons) per cell in a 5 km² raster, disaggregated by vehicle class, driveline, battery capacity and charging strategy.

MOSTACHI only generates simulation results. It is up to the user of the tool to run further analysis of these datasets to answer research questions of interest.

Availability of Code and Data

The MOSTACHI source code and input data used in this publication are available at https://doi.org/10.5281/zenodo.8362266. For the latest version and example output data, please visit https://github.com/JakobRogstadius/MOSTACHI/.

Acknowledgements

This research was funded by Region Blekinge and the European Regional Development Fund through the project “Genomförbarhetsstudie av elvägspilot E22”.

All authors declare that they have no conflicts of interest.

European Scientific Advisory Board on Climate Change. Scientific advice for the determination of an EU-wide 2040 climate target and a greenhouse gas budget for 2030–2050. (Publications Office, 2023).
Regulation (EU) 2021/1119 of the European Parliament and of the Council of 30 June 2021 establishing the framework for achieving climate neutrality and amending Regulations (EC) No 401/2009 and (EU) 2018/1999 (‘European Climate Law’). OJ L vol. 243 (2021).
Buysse, C. & Miller, J. Transport could burn up the EU’s entire carbon budget. International Council on Clean Transportation https://theicct.org/transport-could-burn-up-the-eus-entire-carbon-budget/ (2021).
Mulholland, E. & Rodríguez, F. An analysis on the revision of Europe’s heavy-duty CO2 standards. (2022).
Anderhofstadt, B. & Spinler, S. Preferences for autonomous and alternative fuel-powered heavy-duty trucks in Germany. Transportation Research Part D: Transport and Environment 79, 102232 (2020).
Determinants of Electric Road System Adoption by Road Freight Companies | International Journal of Innovation and Technology Management. https://www.worldscientific.com/doi/10.1142/S0219877023500323.
de Saxe, C. et al. An Electric Road System or Big Batteries: Implications for UK Road Freight. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4194379 (2022).
Deshpande, P., de Saxe, C., Ainalis, D., Miles, J. & Cebon, D. A breakeven cost analysis framework for electric road systems. Transportation Research Part D: Transport and Environment 122, 103870 (2023).
Börjesson, M., Johansson, M. & Kågeson, P. The economics of electric roads. Transportation Research Part C: Emerging Technologies 125, 102990 (2021).
Coban, H. H., Rehman, A. & Mohamed, A. Analyzing the Societal Cost of Electric Roads Compared to Batteries and Oil for All Forms of Road Transport. Energies 15, 1925 (2022).
Shoman, W., Karlsson, S. & Yeh, S. Benefits of an Electric Road System for Battery Electric Vehicles. World Electric Vehicle Journal 13, 197 (2022).
Craglia, M. Decarbonising Europe’s Trucks: How to Minimise Cost Uncertainty. https://www.itf-oecd.org/decarbonising-europes-trucks-minimise-cost-uncertainty (2022).
Márquez-Fernández, F. J., Bischoff, J., Domingues-Olavarría, G. & Alaküla, M. Assessment of Future EV Charging Infrastructure Scenarios for Long-Distance Transport in Sweden. IEEE Transactions on Transportation Electrification 8, 615–626 (2022).
Domingues-Olavarría, G., Márquez-Fernández, F. J., Fyhr, P., Reinap, A. & Alaküla, M. Electric Roads: Analyzing the Societal Cost of Electrifying All Danish Road Transport. World Electric Vehicle Journal 9, 9 (2018).
Nordin, L., Wehner, J., Käck, S. & Brunner, S. Nyttjandegrad elväg : omvärldsanalys och faktorer som påverkar nyttjandegraden av elvägar. (Statens väg- och transportforskningsinstitut, 2023).
Lindgren, M. Electric road system technologies in Sweden. (2020).
Andersson, M. et al. Regulating Electric Road Systems in Europe - How can a deployment of ERS be facilitated? A discussion paper from the CollERS2 project. (2022).
Fynn, C. & Knezevic, G. European launch vision for Electric Road Systems. https://usercontent.one/wp/www.ikem.de/wp-content/uploads/2022/11/Studie-European-launch-vision-for-Electric-Road-Systems.pdf?media=1667839188 (2022).
Rogstadius, J. Interaktionseffekter mellan batterielektriska lastbilar, elvägar och statisk laddinfrastruktur : Resultat från högupplöst simulering av godstransporter på det svenska vägnätet under perioden 2020–2050. (2022).
CEDR Working Group 3.5 Performance. Trans-European Road Network, TEN-T (Roads): 2019 Performance Report. https://www.cedr.eu/download/Publications/2020/CEDR-Technical-Report-2020-01-TEN-T-2019-Performance-Report.pdf (2020).
Basma, H., Zhou, Y. & Rodriguez, F. Fuel-cell hydrogen long-haul trucks in Europe: A total cost of ownership analysis. http://theicct.org/publication/eu-hvs-fuels-evs-fuel-cell-hdvs-europe-sep22 (2022).
Plötz, P. Hydrogen technology is unlikely to play a major role in sustainable road transport. Nat Electron 5, 8–10 (2022).
Granovskii, M., Dincer, I. & Rosen, M. A. Economic and environmental comparison of conventional, hybrid, electric and hydrogen fuel cell vehicles. Journal of Power Sources 159, 1186–1193 (2006).
Kim, H., Hartmann, N., Zeller, M., Luise, R. & Soylu, T. Comparative TCO Analysis of Battery Electric and Hydrogen Fuel Cell Buses for Public Transport System in Small to Midsize Cities. Energies 14, 4384 (2021).
Offer, G. J., Howey, D., Contestabile, M., Clague, R. & Brandon, N. P. Comparative analysis of battery electric, hydrogen fuel cell and hybrid vehicles in a future sustainable road transport system. Energy Policy 38, 24–29 (2010).
Alfredsson, H. & Rogstadius, J. Ruttbaserade simulerade trafikdataför högupplöst analys av tunga godstransporter på det svenska vägnätet. (2022).
Trafikverket. Analysera och prognosticera godstrafik : Samgodsmodellen förklarad på ett enklare sätt. (2017).
Hicks, J. R. The Foundations of Welfare Economics. The Economic Journal 49, 696–712 (1939).
Trafikverket. Analysmetod och samhällsekonomiska kalkylvärden för transportsektorn: ASEK 7.0: Kapitel 14 Operativa trafikeringskostnader för godstransporter. https://bransch.trafikverket.se/for-dig-i-branschen/Planera-och-utreda/Planerings--och-analysmetoder/Samhallsekonomisk-analys-och-trafikanalys/gallande-forutsattningar-och-indata/ (2020).
Plötz, P. & Speth, D. Truck Stop Locations in Europe. 33 https://www.isi.fraunhofer.de/content/dam/isi/dokumente/cce/2021/ACEA_truckstop_report_update.pdf (2021).
Speth, D., Sauter, V., Plötz, P. & Signer, T. Synthetic European road freight transport flow data. Data in Brief 40, 107786 (2022).
Öppna geodata för arbetsplatsområden utanför tätort. Statistiska Centralbyrån https://www.scb.se/vara-tjanster/oppna-data/oppna-geodata/verksamhetsomraden/.
Project OSRM. https://project-osrm.org/.

There is NO Competing Interest.

supplementarymaterial.pdf

Download PDF

Version 1

posted

You are reading this latest preprint version

Electric Road Systems: A No-Regret Investment with Policy Support

Status:

Version 1

Abstract

Figures

Introduction

Limitations

Experimental Design

Simulation Model

Scenario-Based Robustness Analysis

Results

System Cost and Fossil GHG Emissions

Charging Infrastructure Return on Investment

Vehicles and Energy Carriers

Battery Demand

ICE Fuel Demand

Charging Infrastructure Demand

Conclusions

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1