Estimating stocks and flows of electric passenger vehicle batteries in the Norwegian fleet from 2011 to 2030

Retired passenger battery electric vehicles (BEVs) are expected to generate significant volumes of lithium‐ion batteries (LIBs), opening business opportunities for second life and recycling. In order to evaluate these, robust estimates of the future quantity and composition of LIBs are imperative. Here, we analyzed BEV fate in the Norwegian passenger vehicle fleet and estimated the corresponding battery capacity in retired vehicles from 2011 to 2030, using a stock‐flow vehicle cohort model linked to analysis of the battery types and sizes contained in different BEVs. Results based on this combination of modeled and highly disaggregated technical data show that (i) the LIB energy capacity available for second use or recycling from end‐of‐life vehicles is expected to reach 0.6 GWh in 2025 and 2.1 GWh in 2030 (not accounting for any losses); (ii) most LIBs are currently contained within the weight segment 1500–1599 kg followed by 2000+ kg; (iii) highest sales currently exist for BEVs containing lithium nickel manganese cobalt oxide (NMC) batteries; and (iv) lithium nickel cobalt aluminum oxide batteries initially constitute the largest overall capacity in retired vehicles, but will later be surpassed by NMCs. The results demonstrate rapidly growing opportunities for businesses to make use of retired batteries and a necessity to adapt to changing battery types and sizes.

currently the most mature zero emission technology in use, relying primarily on lithium-ion batteries (LIBs). Between 2011 and 2019, Norwegian passenger battery electric vehicle (BEV) sales rose from approximately 2000 to 60,345, with BEVs representing about 42% of the passenger vehicle market in 2019 (OFV, 2019). This represents one of the highest market shares worldwide (IRENA, 2017). Consequently, Norway is expected to be one of the first countries with a significant number of retired batteries (Casals et al., 2017), giving rise to major opportunities that include recycling with material recovery (Velazquez-Martinez et al., 2019) or second use as stationary energy storage applications (Ahmadi et al., 2017;Cusenza et al., 2019).
Recycling and second use of BEV batteries is already ongoing, albeit with a relatively low number of end-of-life BEV inflows. Information from "Batteriretur," a Norwegian company responsible for collection and treatment of used batteries, reveals that Norwegian BEV LIBs are generally collected and dismantled to module level in Norway before export to the European Union (EU) for recycling (Svendsen, T. H., personal communication, September 14, 2020). The recycling process is mainly focused on thermal pretreatment before crushing, or batteries may be refurbished/repaired and re-used in vehicles or for other second uses. However, this is dependent on levels of degradation and other faults. Since end-of-life volumes of BEV LIB are currently small, most batteries currently collected derive from accidents and take-back campaigns, but volumes are expected to increase rapidly in the next decade as the market share increases, sales rise, and vehicles retire (Hao et al., 2015;Palencia et al., 2012;Richa et al., 2014;Sato & Nakata, 2020). Quantitative information about the expected future development of retired batteries and an understanding of their drivers is needed to grasp these opportunities, for example, for planning investments in recycling or reuse infrastructures.
Dynamic stock modeling, including material flow analysis, has been used to assess the development of future electric vehicle fleets and forecast end-of-life vehicle and battery flows (Hao et al., 2015;Palencia et al., 2012;Richa et al., 2014;Sato & Nakata, 2020). These models can be based on cohorts where each cohort is assigned an expected lifetime and the cohort's use phase ends when its lifetime elapses. Using sales scenarios and a discrete lifetime distribution for batteries, Bobba et al. (2019) estimated that a total of around 450,000 battery units will leave the European fleet in 2030, and that under two scenarios with low and high second use the actual battery capacity available for second use will be 1. 99 and 8.75 GWh (70,400 and 311,500 units), respectively. Other studies based on sales scenarios combined with typical vehicle exit curves or average battery lifetimes, respectively, conclude more conservatively that a total of 125,000 electric vehicles (EV) and the batteries they contain will be scrapped in 2030 in Europe (Element Energy, 2019), or less conservatively that a total of 1.2 million EV batteries (47 GWh) will reach end-of-life in Europe in 2030 (Drabik & Rizos, 2018). Of the former, the authors expect that 15% of battery units may be sent to recycling due to deterioration, and 2.25 GWh (representing 105,000 batteries) may be available for second life. At the combined Nordic level, Dahllöf et al. (2019) estimated from historical vehicle sales and battery lifetime data that around 50,000 and 20,000 battery units would be available together in 2030 for second life and recycling, respectively, but this only accounts for batteries already placed on the Nordic market in 2018. The wide variation in results reflect variation in scope, system boundaries, and inherent uncertainties.
Even though reuse and recycling opportunities are likely to arise first in Norway, no studies to the authors' knowledge have yet fully quantified the Norwegian battery volumes arising to 2030. In addition, no studies estimating battery capacity in retired vehicles in Europe could be found that are fully based on the historical differentiation of vehicles arriving into the market and the individual technical battery characteristics linked to each vehicle make/model and sales year. Here, in addition to providing new analysis of the state of the art of battery use in Norway, we estimate the quantity of LIBs entering and leaving the Norwegian passenger vehicle fleet annually until 2030. The target is to investigate short-to medium-term potentials for recycling opportunities for Norwegian industries, so a dynamic stock model is consequently used to build realistic scenarios for the battery capacity becoming available for recycling in future years. The strength of our approach is the combination of modeled data with a large amount of real, technical data at a vehicle model level, based on individual battery characteristics of each BEV sold in Norway. The result is a battery capacity stock and flow model specific to Norway, although the approach could further be applied to other regions to explore their own potential for recycling.

METHODOLOGY
Results of a vehicle stocks and flows cohort model based on the Norwegian market were linked together with supplementary battery analysis based on BEV historical data and anticipated battery development. An overview of the model and analysis linkage is shown in Figure 1.

Application of the stocks and flows cohort model
Passenger BEV stocks and flows were projected to 2030 using a previously developed cohort model that accounts for all initial BEV stocks introduced since 1981 when the first registered electric vehicle sales in Norway occurred (L. Fridstrøm, 2019;L. Fridstrøm et al., 2016 In essence, survival rates are calculated by observing the change in the stock of a given cohort of vehicles from one year to the next. There is currently limited data available to calculate the survival rates of passenger BEVs older than 6-8 years, but since there is rapid technological development that means that models soon become outdated, BEV survival rates for each weight segment were set in the model similar but somewhat lower to those of correspondingly sized petrol-driven vehicles. The limited evidence so far suggests that BEV batteries last the life of the vehicles, which is consequently assumed here. Survival curves for different BEV weight segments used in the model, as well as a discussion of assumptions, are shown in Supporting Information S1. Knowing the survival rate of each vehicle segment to the next year, and accounting for secondhand sales of imports, allowed us to estimate annual fleet stock changes for all vehicles older than 1 year. In this way, estimates were made of the change in the number of vehicles from different first registration years and for different weight segments. Equation (1) shows the relationship between the defined annual stock change of these vehicles dS t>1 and the outflows O which aggregates exports, deregistration, and scrappage. This suggests that for small numbers of vehicle imports older than 1 year (which we assume here), the stock change can be set equal to the outflows. Total stock change for the whole fleet (dS total ) can be thereafter calculated by summing up the inflow of vehicle sales and imports of vehicles less than 1 year old (I t<1 ) with the stock change of older vehicles (Equation 2). dS total was not needed for this study, so Equation (2) serves only to demonstrate the difference between dS total and dS t>1 . Finally, Equation (3) shows how the vehicle outflows are calculated using a survival function sf(s) t,c specific for each vehicle segment s, which is applied to the stock S. This function determines the share of vehicles of a given cohort that remain in the fleet at any given time.
The stocks and flows cohort model itself does not make any assumptions about battery characteristics of the vehicles, but this analysis relating to battery quantities was retro-actively performed using the output (see Sections 2.2 and 2.3). Although the weight segment <1000 kg is included in the model as standard, this category was excluded from subsequent analysis since it was assumed that these vehicles in this category are registered as four-wheel motorcycles and not passenger vehicles. Note that "age" in the model is defined as the number of years completed by December 31 from initial registration, rounded upward to the nearest integer. For example, vehicles aged "3 years" in 2021 are those first registered in 2019.
Although the model includes electric vehicles produced from the year 1981, significant LIB BEV annual sales did not occur until after 2010/2011.

Assessment of electric vehicle battery characteristics
Analysis to estimate LIB capacity from the cohort model results was performed based on historical and statistical data of Norwegian vehicle sales (at a vehicle model level), their associated battery characteristics and expected future battery development.
Historical data on all electric vehicle make/model characteristics (including nominal battery capacity, kWh) that have been available on the mar- combinations thereof. Lithium iron phosphate (LFP) has also been used for the <1000 kg segment. In this analysis only overarching battery material types are considered (i.e., NMC is not categorized according to NMC111, NMC622 or NMC811, etc.), due to a lack of reliable and consistent data.
Where no data about battery chemistries was available, vehicle battery types were set to "unknown Li-ion type." Historical sales data of Norwegian passenger BEVs between the years 2011 to 2018 was obtained from Opplysningsrådet for Veitrafikken AS (OFV, 2019). Vehicles <1000 kg were excluded as before. It was also assumed that electric vehicles sold prior to 2011 when the modern BEV was launched were either not of LIB type, or were registered as four-wheel motorcycles, and were excluded. The sales data was thereafter combined with the background data of battery type and size for different vehicle makes/models to assign a battery capacity and type to each vehicle sold.
Examples of data for the five most popular passenger BEV models, reflecting around 70% of all vehicles sold in Norway between 2011 and 2018, are shown in Table S2 in Supporting Information S1. The combined historical sales and background battery data was used to estimate the amount of type of batteries introduced into the Norwegian passenger vehicle fleet between 2011 and 2018.
In preparation for combination with the stocks-flows cohort model results, the sales of passenger BEVs and associated battery characteristics were grouped into the same weight segments as for the cohort model by using associated vehicle curb weights in the EV database (and accounting for a 75 kg driver). The data was also transformed to calculate the sales weighted average battery capacity and type for Norwegian passenger BEVs purchased in each weight segment and for each vehicle sale year. Where several battery types were used for vehicles sold within one weight segment (and for one vehicle sale year), a weighting factor was determined to estimate the distribution of vehicles actually sold according to battery type. Any gaps in weight segments/years were filled with data from an adjacent weight segment, and data for the year 2019 was assumed the same as 2018.
The estimated battery characteristics were extended to 2030. Although the Electric Vehicle Database also contains the available information about known models arriving to the market in future years (to 2022), few models beyond 2021 have been announced and there is thus little concrete information available about the growth in battery capacity to 2030. Within each segment there is a band of battery capacities; we therefore assumed for this analysis that the maximum capacity in each segment will continue to increase and that the sales weighted average battery capacity will converge toward the upper end of these bands in all segments by 2030, with the phasing out of older vehicle models and the demand for longrange driving. We also assume that large and luxury vehicles will develop an even larger battery capacity, in the region of 90-120 kWh. Resulting assumptions of battery capacity growth used in the analysis here to 2022-and beyond to 2030-are shown in Table S1 in Supporting Information S1, with linear approximation used to extend current capacity values from today. Due to a lack of reliable data on the battery types of future models, battery types for years 2020-2030 were set to unknown Li-ion.

Estimation of new batteries and stock change annually until 2030
The number and capacity of batteries of different types entering the electric passenger vehicle fleet, as well as the stock change, were estimated for Model uncertainties (1) originate from the fact that only one scenario of BEV penetration was investigated, and that the modeled stock change of vehicles older than 1 year (i.e., excluding new vehicle sales) was assumed to equate to scrappage. In reality the stock change of these vehicles is also affected by imports and exports, as well as other contributions from deregistration, but these individual flows are not estimated by the model. To establish how the import/export flows may affect total vehicle outflows estimated by the stocks-flows model, these flows were investigated further using data from the year as an example (SSB, 2020a).
Uncertainties in battery capacity development (2) also affect results, reflecting underlying complex dynamics beyond the scope of this study.
For example, as technology advances and batteries become more efficient, several trends can unfold. First, the efficiency gains can be used to further increase battery capacity and driving range. However, this may be limited in the medium, compact, and smaller vehicles compared to larger vehicles due to costs (potentially exacerbated by constraints in raw material and battery supply) and improvements in charging infrastructure.
Second, efficiency gains can be used to reduce the battery size. This option would reduce battery and vehicle weight and consequently also increase the range while keeping costs low, but could lead to the stagnation of the maximum battery capacity.

Application of the stocks and flows cohort model
The total Norwegian fleet of passenger BEVs to 2030 based on the Norwegian National budget, estimated by the stocks and flows cohort model, is shown in Figure 2. Up to and including 2018, actual data on the number of vehicles of different technologies that have been registered each year has been used, based on data from the national vehicle register.
Annual results from the model of total new passenger BEV sales, and stock change of older vehicles (age > 1 year), are shown in Figure 3 are approximately 10-25% higher than new vehicle sales registered by OFV, but when the number of new registrations from secondhand imports registered by OFV is also considered (as also implemented in the cohort model), than the difference is <4%. See Table S3 in Supporting Information S1 for more details. Many of these latter vehicles have already been registered abroad once before during the same year and have been imported secondhand due to the high demand for some popular models in Norway that have not been available in sufficient volumes. The fact that BEVs are subsidized in many countries, but not directly in Norway, gives rise to business opportunities particularly within secondhand import. In some cases, vehicles have been registered in an EU country for just a day to be counted toward the EU CO 2 requirement before being exported to Norway. This per year of passenger BEVs older than 1 year, that is, the net stock change of the older vehicles that were already in the fleet each year, excluding new vehicle sales that year. Since we assume here that imports of older vehicles are negligible, this equates to fleet outflows due to scrappage, deregistration or export. The numbers also directly equate to the number of battery packs in these vehicles (i.e., one per vehicle).
For this article, the assumption is that vehicles in the outflows are mostly scrapped in Norway rather than exported. Historically this has been the case due to the high taxes on passenger vehicles compared to other countries, which make old used vehicles more valuable in Norway than in other countries. Since BEVs do not have purchase taxes in Norway, they could potentially be exported to other countries. However, the user demand for BEVs has been much higher in Norway than elsewhere, which makes it reasonable to assume these flows to be negligible. For battery electric trucks and buses the situation may be different. For verification, comparisons were made of the total net vehicle stock change estimated by the model for all vehicle types and ages with historical scrappage data from years 2010 to 2018 (SSB, 2019). Results, shown in Figure 3c, are comparable.
Whilst inferring that other flows contributing to the stock change for these older vehicles are small in comparison to scrappage, the data reflects the situation for the entire vehicle fleet and not specifically for BEVs. This is since scrappage data specifically of passenger BEVs in Norway is not publicly available for detailed comparisons.

Effects of imports and exports on estimated outflows
It was assumed for this work that imports of older vehicles than 1 year, and exports of all ages, are negligible, which makes the stock change (vehicle age > 1 year) equate to outflows (cf. Equation 1). These assumptions are investigated here in more detail. Figure 4 shows estimated outflows from the stocks and flows cohort model broken down by vehicle age. For 2015 and 2020 a significant fraction of the outflow is constituted by vehicles younger than 7 years. This is expected, since the majority of EVs have not yet reached end-of-life and therefore the main cause of outflows are accidents, callbacks, or malfunctions of any nature. For 2025 and 2030 these outflows make up for a smaller share and the main outflow of BEVs is around 10 years old. Although these vehicles are still short of their full lifetime, the reason for this trend lies in the relative differences in cohort abundance: BEVs aged 10 years are still the most scrapped in 2030 because they are more numerous than older vehicles. Correspondingly, even if their scrappage rate is low, the absolute number of those scrapped is higher than for older vehicles. Differences in the spread of vehicle ages can also be seen in the figures. In 2015 many older vehicles (dating back to 1981) were phased out due to the rapid market development. Between 2020 and 2030, the spread of vehicle outflow age is anticipated to widen as time increases from 2011 when the rapid BEV introduction began, and the vehicles are able to progress along their survival curves. The majority of imported/exported vehicles can be assumed to be relatively young, for example, less than 5 years old, and hence they are unlikely to significantly affect the main outflows shown in Figure 3b. However, to establish how import/export flows may affect total vehicle outflows, these flows were investigated further using the year 2018 as an example. For this year, total recorded imports of new and used vehicles to Norway equaled 50,840 and 11,913, respectively, whilst exports of new and used vehicles from Norway were much lower at 10 and 46, respectively (SSB, 2020a).
This imbalance is not unexpected and is due to the subsidies paid out in many countries that make it profitable to import BEVs into Norway coupled with high demand in Norway compared to other countries.
Since export flows of BEVs are almost negligible and imports dominate, the calculations for BEV scrappage, and associated estimates of the batteries they contain, may be underestimated. However, most imported vehicles to Norway are likely to be nearly new (age < 1 year) and were therefore accounted together with new BEV sales in the model. Recorded data shows that for the year 2018, there were 11,899 first time registrations of imported vehicles in Norway (OFV, 2020). Although these do not necessarily derive from the total pool of 11,913 used vehicles imported during 2018 (vehicles can also derive from previous year imports), the difference is small. Since vehicles of age > 1 year are directly counted along with new sales in the stocks-flow cohort model as "new vehicles," it is unlikely that used imports have a large impact on the estimates of vehicle scrappage in this study.

Assessment of electric vehicle battery characteristics
Data of the development in battery capacity for all vehicles available on the market, including BEVs known to be arriving on the market in the next years, is shown in Figure 5. Both the average and maximum battery capacity of BEVs available on the market per year has in general shown an upward growth trend since BEV introduction, although the growth can in most cases be described as stepwise. Little is known about models arriving on the market after 2021, aside from several examples in the 1400-1499 kg and >2000 kg segments. For the latter, the large increase in maximum capacity relates to the announcement of the new Tesla Roadster, anticipated in 2022, with 200 kWh battery capacity per vehicle. However, this is unlikely to be representative of the whole segment.
Estimates of the types of batteries entering the fleet based on historical sales data combined directly with known battery characteristics for these vehicle models are shown in Figure 6. According to these results, NMC and NCA are battery types currently used in greatest amounts, with around 0.9 and 0.7 GWh entering the fleet in new passenger BEV sales in 2018, respectively (Figure 6a). There is also a division of battery types by packs entering the fleet coupled with an increase in battery size. As technology improves it can be expected that lighter batteries will be able to deliver the same energy capacity, resulting in a positive rebound effect in which fewer materials are required to provide the same service. The range of modern BEVs is already approaching that of ICE vehicles, suggesting that further developments will soon focus on reducing battery sizes and therewith EV prices.
There is large uncertainty regarding future battery chemistries, but an overall trend to move away from cobalt seems to be dominant throughout the industry as can be seen by efforts to move from NMC111 to NMC811 (Alves Dias et al., 2018;Azevedo et al., 2018). Tesla, the main NCA battery user, has also expressed commitment to reducing cobalt use through increased use of nickel and, as has been seen in the Chinese market, moving toward LFP batteries (Holland, 2020). While this trend strengthens raw material supply security, it may result in problem-shifting toward scarcity of nickel supply.

Estimation of new batteries and net change annually until 2030
Output from the stocks and flows cohort model was combined with the supplementary battery analysis to estimate the respective battery flows until 2030. Results are shown in Figure 7, with an in-depth summary of annual net stock change for the years 2017-2025 given in Figure S3 in Supporting Information S1 that represents the arising Norwegian "window of opportunity" for end-of-life BEV, for both recycling and second-life purposes. The large increase in battery capacity entering the fleet between 2019 and 2022 is due to the increase in assumed battery sizes in many weight classes to 85% of their 2030 value, as described in Table S1 in Supporting Information S1. According to results, total battery amount used in  Walz, 2018). Together these companies have announced the formation of a joint venture to enable recycling of battery materials and aluminum from electric vehicles by building a pilot battery recycling plant, which will be the first of its kind in Norway (Hydro, 2020). The quantity of assumed end-of-life batteries estimated here represents the potential total available for both recycling and second-life concepts. If the quantity of batteries assumed going mostly to scrap is instead allocated wholly for second-life purposes, these batteries could potentially feed 70,000 and 260,000 typical home/cabin battery energy systems of 8 kWh in 2025 and 2030, respectively (Alternativ Energi AS, 2020). Nevertheless, far from all batteries can be re-used due to degradation or other faults (Svendsen, T. H., personal communication, September 14, 2020). These applications are still in an early phase and the dynamics will depend on a range of factors from policy to business opportunities. Based on the recently released EU proposal for the battery legislation it seems that incentives are targeted toward recycling, while reuse will be rather market regulated. Reuse can be seen as delaying the availability of secondary raw materials for automakers, while potentially also reducing the demand for new batteries for stationary applications. Thus, there is a need to further study these dynamics and better understand the impact of reuse and recycling for material security.
Complicating the picture, differences in recycling and reuse economic viability are relatively unknown at present, and other types of losses will also affect the actual total quantity of batteries available for recycling and second life. It is assumed in many studies that around 10% of vehicles are lost and not collected when scrapped, and there is widespread criteria established in the literature for EV battery retirement that capacity is reduced to 70−80% at first end-of-life (Martinez-Laserna et al., 2018;Saxena et al., 2015;Zhao, 2017

CONCLUSIONS
Short-to medium-term potentials for recycling opportunities for Norwegian industries were investigated in this study; the total number of battery packs in new passenger BEV sales in Norway was estimated to be 116,000 in 2025 and 163,000 in 2030, and the number in retired vehicles to be approximately 17,000 in 2025 and 51,000 in 2030. In terms of battery capacity, this equates in new sales to 2.1 GWh in 2025 and 11.0 GWh in 2030, and in retired vehicles, 0.6 GWh in 2025 and 2.1 GWh in 2030 (not accounting for losses). Results show that NMC and NCA are battery types currently used in greatest amounts, and that there is also a division by weight segment evident. Most LIBs are currently contained within the weight segment 1500-1599 kg followed by the weight segment 2000+ kg. NCA is in use for heavier weight segments and LMO/NMC in use for lighter weight segments. In terms of LIB types in retired vehicles, NCA batteries initially constitute the largest overall capacity, but we estimate they will be surpassed by NMCs in later years. Not included in calculations are batteries from plug-in hybrid electric vehicles (PHEVs) and batteries from battery electric light commercial vehicles (BE-LCVs). However, these constitute lower fleet vehicle volumes; at end of year 2019 there were 116,042 PHEVs and 7332 BE-LCVs versus 260,292 BEVs (SSB, 2020b), with PHEVs also having smaller battery capacity than BEVs per vehicle.
Since the study builds on multiple modeling processes, various uncertainties are present. Although an overall trend to move away from cobalt seems to be dominant throughout the industry, very little concrete data is publicly available about the specific type of Li-ion batteries future BEV models will utilize. Thus all batteries arriving into the fleet between 2020 and 2030 were assigned in this study as unknown Li-ion type. This simplification allows the forecast uncertainty to be reduced but leaves unanswered questions about the end-of-life materials available. Other key uncertainties relate to the lack of differentiation of import and export flows in the stocks-flow model output, and the non-inclusion of other detailed types of outflows (e.g., vehicle and battery capacity losses), that will also affect the main results. For the former uncertainty, the available data suggest that exports are currently low and the majority of used vehicles imported in recent years are less than 1 year old, which significantly reduces the model uncertainty. Nevertheless, this may change over time.
In summary, this analysis based on a combination of vehicle-specific data and assumptions of BEV market uptake from the Norwegian national budget estimates that the battery capacity and pack number in retired BEVs will increase dramatically toward 2030, indicating great potential for domestic markets to develop around battery recycling and reuse. Further, it provides insights into the materials embedded in the batteries as well as a theoretical framework that can be applied to other regions. The results also indicate that it will be necessary to adapt to changing battery types and sizes of the retired batteries. Making use of business opportunities activities will require a large amount of infrastructure, as well as new regulations, for which the estimates provided here can act as a guide.

ACKNOWLEDGMENTS
This work was carried out in the BATMAN project funded by the Research Council of Norway and partners (NFR: BATMAN-a299334). Limited text in this article is taken from an earlier, related report (Figenbaum et al., 2020) prepared by several of the authors. Thanks are due to BATMAN coordinator Stephen Sayfritz (Eyde Cluster), for constructive feedback at all stages. Special thanks are also due to Tor Henrik Svendsen (Batteriretur) for Norwegian state-of-the-art battery processing information.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from Opplysningsrådet for veitrafikken (OFV) and EV Database. Restrictions apply to the availability of these data, which were used under license for this study. Data are available directly from OFV and EV Database.