The emergence of cost effective battery storage

The cost of energy storage

The primary economic motive for electricity storage is that power is more valuable at times when it is dispatched compared to the hours when the storage device is charged8,12,16,17,18. These benefits will accrue over the entire lifetime of the storage system and must be weighed against the cost of acquiring a system capable of performing the storage service for a given number of charging/discharging events per year over the useful life of the system. A battery will be sized in the two dimensions of power and energy capacity. The size of the power component, measured in kW, governs the maximum rated electricity charge/discharge rate. The energy component determines the total capacity of electricity that can be stored. It is measured in kWh. Moreover, the ratio of energy capacity to rated power determines the duration for which the storage facility can provide the rated power. This is also the length of time needed to charge the battery given its power rating.

To capture the unit cost associated with energy storage, we introduce the Levelized Cost of Energy Storage (LCOES) which, like the commonly known Levelized Cost of Energy, is measured in monetary units (say U.S. $) per kWh. Similar to the LCOE that indicates the average revenue an investor would need in order to break-even over the life cycle of a power generating facility19, the LCOES measure captures the break-even value for charging and discharging electricity on a per kWh basis.

Earlier studies on the cost of storage have usually fixed the duration of the storage system at some exogenous value, for instance, 4 h18,20. In contrast, we first decompose the overall unit cost into the levelized cost of energy components, LCOEC (in $ per kWh), and the levelized cost of power components, LCOPC (in $ per kW). As shown in the Methods section, these levelized costs are obtained by dividing the system price of the power and energy components, respectively, by the total discounted number of charge/discharge occurrences that the battery performs the storage service in the course of its useful life. In particular, the number of charge/discharge events per year is multiplied by a factor that reflects the useful life of the battery, the cost of capital (discount rate), round-trip efficiency losses and, finally, performance degradation over time. It is presumed that these cycling occurrences are not limited to only full capacity charging and discharging, rather, they would include partial charge/discharge events where the capacity of the battery is not fully charged or discharged.

For a storage system with a power rating of kp kW and a storage capacity of ke kWh, the corresponding average duration is defined as \(D \equiv \frac{{k_{\mathrm{e}}}}{{k_{\mathrm{p}}}}\) hours. The duration, D, indicates the number of hours that the system charges and discharges kp kW of power per full cycle. On a lifetime basis, the cost of storing one kWh of electricity, and dispatching it at later hours of the same cycle is LCOES (D) = LCOEC + LCOPC · \(\frac{1}{D}\). Since D is stated in hours, LCOES(⋅) is expressed in $ per kWh. The following claim identifies the LCOES metric as the break-even price per kWh for electricity storage services.

Claim: \({{\mathrm{LCOES}}\left({\frac{{k_{\mathrm{e}}}}{{k_{\mathrm{p}}}}} \right) = {\mathrm{LCOEC}} + {\mathrm{LCOPC}} \cdot \frac{{k_{\mathrm{p}}}}{{k_{\mathrm{e}}}}}\) is the break-even price for storing and dispatching ke kWh of energy in N charging/discharging events per year, subject to the maximum power charge (discharge) not exceeding kp kW at any point in time.

The preceding claim is formally validated in the Methods section below. To calibrate the LCOES metric in the context of lithium-ion, batteries, the following calculations are based on current U.S. market prices of $171 per kWh and $970 per kW for energy and power components, respectively, in the context of small residential applications (see Supplementary Figures 1, 2 and Supplementary Tables 1–3). In the context of lithium-ion batteries, we expand the cost model in order to allow for certain costs related to installation to be entirely independent of the size of the battery, e.g., permitting, inspecting,and commissioning. In the location-specific application of our model, these fixed costs are estimated to be $400 in the U.S. and $300 (€260) in Germany (see Supplementary Notes 1–6).

Assuming N = 365 charging/discharging events, a 10-year useful life of the energy storage component, a 5% cost of capital, a 5% round-trip efficiency loss, and a battery storage capacity degradation rate of 1% annually, the corresponding levelized cost figures are LCOEC = $0.067 per kWh and LCOPC = $0.206 per kW for 2019. The solid curve in Fig. 1 shows the corresponding LCOES for alternative duration values. In the presence of installation related fixed costs, LCOES then yields the break-even price for covering the systems costs that do vary proportionally with either ke or kp; see Methods section for further details.

Simulated trajectory for lithium-ion LCOES ($ per kWh) as a function of duration (hours) for the years 2013, 2019, and 2023. For energy storage systems based on stationary lithium-ion batteries, the 2019 estimate for the levelized cost of the power component, LCOPC, is $0.206 per kW, while the levelized cost of the energy component, LCOEC, is $0.067 per kWh. The curve corresponding to the year 2019 plots the corresponding LCOES values for alternative levels of the storage system’s duration. The LCOES curve corresponding to the year 2013 indicates the decline in lithium-ion based battery storage costs over the past five years. The 2023 curve projects anticipated future cost reductions. Source data are provided as a Source Data file

Full size image

Consistent with the recent widespread installation of li-ion based batteries, the LCOES of such systems has dropped dramatically in recent years21,22,23. The dotted curve shows the corresponding nominal LCOES figures back in 2013. Projecting into the future, the consensus forecast that emerges from various literature sources are annual percentage declines of 5.6% and 8.1% for the acquisition cost of the power and storage components, respectively, over the horizon 2018–2023. Assuming that this rate of improvement can indeed be maintained on average during those years, the dashed curve in Fig. 1 provides a forecast of where the LCOES of li-ion battery systems is expected to be in 2023 (see Supplementary Notes 2 and 4).

Our LCOES metric is a variant of existing storage cost measures18,20,24,25,26,27. For energy generation, the familiar LCOE measure is frequently conceptualized as total (discounted) cash flows spent divided by total (discounted) energy delivered27,28. Existing studies on the levelized cost of storage follow the same total-cost-divided-by-total-energy approach20,26,29,30. While our LCOES measure is also calibrated as a break-even measure, our metric departs from two individual levelized cost measure (power and energy) and then aggregates these two measures depending on the average duration of the system. This disaggregation will prove useful in characterizing optimally sized storage systems. In particular, the following section shows that the conditions for an optimally sized battery can be expressed succinctly in terms of the optimized duration. By optimizing the duration of the battery storage system, we obtain cost figures that are consistent with the recent widespread and increasing deployment of such storage systems. Earlier studies that arrived at substantially higher cost of storage have frequently fixed the duration at 2 or 4 h20,26. It should also be noted that our LCOES concept only captures the cost per kWh of warehousing electricity on a daily basis, subject to the system’s power rating constraint. In contrast, some earlier studies also include the cost of generating the energy that is being dispatched26,29.

Optimal battery size supplementing a solar PV system

We consider a representative household that has already installed a solar PV system and now faces the question whether behind-the meter storage adds additional value. To illustrate the basic economic tradeoff, we first consider a household for which both the consumption load and the solar generation profile are fairly constant across the seasons of the year. The solid curve in Fig. 2 depicts the average load profile and the bell-shaped dotted curve depicts the average solar PV generation curve.

Pattern of daily charging and discharging of a battery supplementing a PV system. Region I represents self consumption from solar generation; region II is surplus energy that can be stored and subsequently discharged as region IV (minus efficiency losses); and region III is surplus energy sold to the grid. Region V is residual demand that would not be met by the battery and must be met through purchases from the grid at the going retail rate p

Full size image

Absent any battery storage, the household will self-consume the energy represented by the area marked I in Fig. 2, and buy the energy outside the time interval [t−, t+] at the going retail rate, denoted by p. The surplus energy from the solar system, i.e., the regions marked II and III, can possibly be sold back to the energy service provider at some overage tariff, OT which, in Germany, is given by the prevailing feed-in tariff. If a battery system is added, the energy corresponding to the region marked as II in Fig. 2 would be discharged during times when household demand exceeds generation by the rooftop solar facility. Accordingly, region IV in Fig. 2 is equal to region II minus the round-trip efficiency losses. The following derivation also maintains the implicit assumption that the demand represented by combined areas represented IV and V is sufficiently large to absorb the stored energy corresponding to II. Region V is residual demand that would not be met by the battery and must be met through purchases from the grid at the going retail rate p. The general case—which underlies our calculations and empirical findings—is presented in the Methods section.

A battery supplementing an existing solar rooftop system will add value only if the difference between the retail rate p and the overage tariff, OT, is sufficiently large to cover the levelized cost of the optimally sized battery. With a higher power rating, kp, the amount of energy that can be stored will grow at a diminishing rate (Fig. 2). Referring to the power generation curve as G(⋅) and the household load curve as L(⋅), the area (energy) corresponding to region II in Fig. 2 becomes:

$$E^ + (k_{\mathrm{p}}) = \mathop {\int}\limits_0^{24} {\left[ {{\mathrm{min}}\{ L(t) + k_{\mathrm{p}},G(t)\} - {\mathrm{min}}\{ L(t),G(t)\} } \right]} dt.$$

(1)

Our characterization of an optimally sized battery follows the standard microeconomic approach of an (household) investor that seeks to maximize the discounted value of future expected cash flows. As shown in Methods, the overall present value of all cash flows associated with the battery storage system (excluding any fixed costs that do not vary with the size of the battery) is proportional to the profit margin, PM, per cycle, given by:

$${\mathrm{PM}}(k_{\mathrm{p}}) = [{\mathrm{pp}} - {\mathrm{LCOEC}}] \cdot E^ + (k_{\mathrm{p}}) - {\mathrm{LCOPC}} \cdot k_{\mathrm{p}}.$$

(2)

Here, pp refers to the price premium, which is the time-averaged difference between p and OT, adjusted for round-trip efficiency losses and the temporal degradation of the energy discharged by the battery. Equation (2) shows that in order for a storage system to add value, the price premium, pp, must exceed the levelized cost of the energy component, LCOEC, yet this is not sufficient because of the need to cover the levelized cost of the power component, LCOPC. The marginal return to systems with a higher power rating is diminishing and therefore a storage system with positive net-present value can be found if, and only if, the daily profit margin is positive for small values of kp. The value of E+(kp) is approximately equal to kp⋅(t+ − t−) for small values of kp (by L’Hospital’s rule). Therefore some storage system will be valuable for the representative household whenever [pp—LCOEC]⋅(t+ − t−) − LCOPC > 0.

To identify an optimally sized battery system in this setting, we refer to \({\cal{I}}_ + (k_{\mathrm{p}})\) as the duration of the marginal power component, for a given power rating, kp, of the battery. Formally, \({\cal{I}}^ + (k_{\mathrm{p}})\) is defined as the length of the time interval(s): I+(kp) ≡ {t∈[0,24]|L(t) + kp < G(t)}. For the battery storage system illustrated in Fig. 2 the duration of the marginal power component is (t* − t*) hours. In general, the duration of the marginal power component is the derivative of the function E+(⋅) with respect to kp. The first-order optimality condition corresponding to (2) therefore is

$${\mathrm{pp}} = {\mathrm{LCOES}}({\cal{I}}^ + (k_{\mathrm{p}}^ \ast )),$$

(3)

with the optimal storage capacity determined by \(E^ + (k_{\mathrm{p}}^ \ast )\) according to Eq. (1). Thus the LCOES concept introduced above can be used to identify the size of the value-maximizing battery system by equating the price premium and the LCOES evaluated at the duration of the marginal power component. To illustrate this criterion in the context of Fig. 2, suppose the investor is considering to add an additional kW of power capacity. That would enable the addition of at most another (t* − t*) kWh of storage capacity because the solar installation shown in Fig. 2 would make at most that many additional kWh of surplus energy available for charging. The incremental revenue of this addition would therefore be pp⋅(t* − t*) kWh, while the incremental levelized cost would be LCOPC + LCOEC ⋅(t* − t*)⋅h. Optimality requires that the incremental cost be equal to the incremental revenue, or after dividing by (t* − t*)⋅h, the optimality condition becomes pp = LCOES ((t* − t*) ⋅ h).

It should be noted that the average duration of the optimal battery storage system exceeds the duration of the marginal power component, because \(\frac{{\hat k_{\mathrm{e}}(k_{\mathrm{p}}^ \ast )}}{{k_{\mathrm{p}}^ \ast }} > {\cal{I}}^ + (k_{\mathrm{p}}^ \ast ).\) This inequality reflects that the optimal duration, \(\frac{{\hat k_{\mathrm{e}}(k_p^ \ast )}}{{k_{\mathrm{p}}^ \ast }}\), is the average of the durations of the inframarginal power components. In Fig. 2, these durations range from the lowest (t* − t*)⋅h to the maximum value (t+ − t−)⋅h.

In Methods we present an extended version of the model that allows for both the solar generation and household electricity demand to vary across the different months of the year. Central to this extension is the variable Es(kp). For a given month (season), s, the energy quantity Es(kp) represents the maximum that can both be charged during the hours of the middle of the day with surplus solar power and discharged in the hours when the household’s load exceeds the available solar rooftop energy. The optimally sized battery will generally be oversized relative to the needs of a particular month and undersized in others. As a consequence, the battery will not be fully charged in certain months and therefore will not go through full charging/discharging events for parts of the year. Similarly, it will depend on the month whether the household is grid-positive in the sense that, on a daily basis, the household sells energy to the grid, yet does not buy power from the grid. In contrast, the battery storage system shown in Fig. 2 leaves the household a prosumer—self-producer and grid-consumer of electricity—on account of regions III and V.

Our analysis is complementary to other studies that have explored residential solar PV coupled with storage. Some studies have sought to calculate the levelized cost of storage with solar PV, though with energy and power components the size of which is exogenously fixed rather than determined through an explicit optimization 20,31. Some studies have determined optimal power and energy dimensions for storage systems of various technologies, though only in the context of a 1-year simulation10. Other studies rely on sensitivity analysis to find optimal solar PV plus storage new-build combined systems without explicit consideration of the dimension of the power component or the corresponding life-cycle cost metric5,16. Finally, some earlier studies have also focused on sizing a storage system when revenues are obtained through price arbitrage27, though in contrast to our approach, these studies do not yield estimates for the optimized levelized cost of energy storage.

Location-specific applications

Incentives for distributed energy generation in Germany have long been provided by feed-in tariffs. For recent solar installations these tariffs have recently been reduced to ≈12 € cents per kWh32. The current retail rates of near 30 € cents therefore create a substantial price premium. In contrast to other jurisdictions around the world, Germany provides only modest direct incentives for battery storage installations33.

Table 1 shows the results of our model for a representative household in Munich. The calculations are based on a model with 12 representative days, one for each month of the year (see Supplementary Table 4 and Supplementary Note 7). For a ks = 6 kWp residential solar installation, we obtain an optimal battery size of \(k_{\mathrm{p}}^ \ast = 0.73\,{\mathrm{kW}}\) and \(k_{\mathrm{e}}^ \ast = 5.1\,{\mathrm{kWh}}\), yielding an average duration of about 7 h. The corresponding levelized cost of storage is LCOES = 8.5 € cents per kWh and, separately, the fixed cost is given as €260.

Full size table

To calibrate our findings, we note that the ratio of ks to \(k_{\mathrm{e}}^ \ast\) aligns with residential battery system sizes observed in Germany34. While our calculations identify the battery capacity yielding the highest net-present value from a pure arbitrage perspective, the household could essentially double the size of the battery and still break-even on the investment. Specifically, the net-present value would still be non-negative if kp = 1.6 kW and ke = 8 kWh. That energy capacity is, in fact, close to the observed average capacity installation in Germany for 201734. Depending on the specific application of the battery system and the frequency of short duration high-power loads, it may be advantageous to opt for a higher power capacity and additional grid purchases.

The optimally sized battery will be fully charged on representative days during five months of the year. Thus \(k_{\mathrm{e}}^ \ast = 5.1 \le E_s(k_{\mathrm{p}})\). In the remaining 7 months there is either insufficient supply of solar power (November–February) or there is insufficient demand in the after sunset hours (June–August). The household will be grid-positive during average days in June, July, and August in the sense that the energy stored in the battery is sufficient to meet the household’s electricity load. Formally, \({\mathrm{min}}\{ k_{\mathrm{e}}^ \ast ,E_s(k_{\mathrm{p}})\} = A_s\) during those months.

Under current rules, solar PV systems in Germany exhaust their feed-in tariff support after 20 years. At that point in time, the household would probably not receive more than the average wholesale rate (around 3 € cents per kWh) for any surplus energy sold back to the energy service provider. For such older solar PV systems, the financial return from adding a battery system would at that point increase further because the price premium would then effectively be at least 27 € cents per kWh. Under this scenario, it would be economical to install a larger power rating (over 4 kW) and coupled larger energy capacity (between 8 €kWh and 11 kWh).

In contrast to Germany, California may at first glance not appear ideally suited for behind-the-meter storage installations. The state’s continued commitment to the policy of net metering effectively allows both residential and commercial customers to sell any temporal surplus electricity back to their energy service providers at the same rate at which the customer acquires electricity from the grid. From the perspective of the customer, this policy effectively enables free energy storage. In the context of the above model, the price premium for storage systems in California would therefore be zero. It should be noted, though, that in confirming the state’s commitment to net metering, California utilities were nonetheless allowed to impose a non-bypassable charge for all electricity consumed by customers with residential solar PV installations. The average non-bypassable charge rate is 2.7¢ per kWh. This surcharge thus becomes an effective price premium because any energy fed back to the grid from the solar facility will only be credited at the basic electricity retail rate.

For new storage installations in California, investors are eligible for both federal and state-level support programs. At the federal level, battery installations in the U.S. qualify for an Investment Tax Credit, ITC, provided the battery can be classified as solar equipment35. Specifically, this requires that the energy storage capability of the battery does not exceed the total energy generated by the solar PV system. As detailed further in Methods, the battery would otherwise only be eligible for a share of the maximum ITC, which amounts to 30% of the acquisition cost.

In addition to the federal support for battery storage systems installed in conjunction with solar PV systems, California has adopted the so-called Self-Generation Incentive Program (SGIP) for behind-the-meter storage systems (see Supplementary Note 8). This program ultimately explains why the majority of residential battery storage systems installed to date in the U.S. are in fact located in California. In its current form, SGIP offers a rebate on the expenditure for energy storage components. Thus it is stated in $ per kWh and effectively reduces the acquisition cost of energy storage components, that is, the parameter ve.

The specific amount to be rebated depends on the duration of the storage system. Normalizing kp at 1 kW, the investor is entitled to a rebate of $400 for the first two kWh of energy storage, an additional rebate of $250 for the next two kWh, and a final rebate of $100 for the next two kWh, up to a duration of 6 h. Additional energy storage components corresponding to the initial 1 kW power rating do not receive any subsidy. For a power component of kp kW, the rebate amounts to $400 for each kWh of energy storage provided the duration of the system does not exceed 2 h36. For systems with longer durations, the rebate per kWh steps down as indicated above, such that no additional support is given to systems with a duration exceeding 6 h. These rebates are available in addition to the federal 30% investment tax credit (ITC) for batteries that qualify as solar equipment.

Table 2 shows the results of our model applied to Los Angeles (see Supplementary Table 5 and Supplementary Note 9). Given a 4.85 kWp residential solar installation, we obtain an optimal battery size of \(k_{\mathrm{p}}^ \ast = 2.45\,{\mathrm{kW}}\) and \(k_{\mathrm{e}}^ \ast = 9.8\,{\mathrm{kWh}}\). The optimal duration is D* = 4 h with a corresponding LCOES of 0.6¢ per kWh and, separately, the fixed cost is given as $400.

Full size table

In contrast to our findings for Germany, the optimally sized battery in California would be charged to capacity effectively every month other than December and January (there is only a small amount of slack of 0.02 kWh in July). At the same time, the representative household will be grid-positive only in the month of June, as in every other month the household will have to continue purchasing power after acquisition of the optimally sized battery. Since the effective price premium is 2.7¢ per kWh, such a system creates value for the investing party under current conditions. Absent any federal or state-level incentives, the LCOES would be 12.8¢ per kWh. The ITC covers 3.8¢(≈0.3 · 12.8¢), while SGIP contributes the remaining 8.4¢.

The fact that the optimal duration in California is exactly 4 h is not a coincidence. It rather reflects that under SGIP the incremental rebate for systems with a duration exceeding 4 h drops from $200 to $100 per kWh. In conjunction with the federal ITC, the California rebate is sufficiently large so as to incentivize a duration between 4 and 6 h, even in the absence of any price premium. The only reason a California household would not install an arbitrarily large storage system is that the full federal tax credit would no longer be available once the energy storage capability exceeds 12.2 kWh, which is the annual average daily production from the solar installation.

We note that an implication of the model is that SGIP could result in economic storage systems (i.e., PM(kp) ≥ 0) with an oversized power rating, relative to what is needed to store the peak production surplus from solar generation. The largest system that would be economical to install has the dimensions of kp = 2.9 kW and ke = 11.6 kWh. In order to maintain a relatively short duration and thereby qualify for a correspondingly high rebate under SGIP, the household will acquire a power component that is 20% larger than optimal. To witness, the duration of the optimally sized battery in California is about half of that in Germany. This conclusion needs to be qualified by the observation that our analysis has employed average hourly load data that do not reflect short-term spikes in power consumption. If such spikes occur with sufficient frequency (e.g., electric vehicle charging, simultaneous clothes drying, and air conditioning operation), the household could be left with a tradeoff between greater autonomy resulting from a more powerful battery and additional electricity purchases from the grid.

Researchers at the US Department of Energy's National Renewable Energy Laboratory (NREL) have assessed the cost and performance of most long-duration energy storage (LDES) technologies. They have also looked at flexible power plants to help electricity systems to deal with extremely high levels of renewable energy penetration and have found that, given current and future capital cost scenarios, that geologic hydrogen storage and natural gas combined-cycle (NGCC) plants with carbon capture storage (CCS) technologies offer the lowest levelized cost of energy (LCOE) for 120-hour discharge applications and that pumped hydro, compressed air, and batteries are the cheapest solutions for 12-hour discharge.

“Since energy storage technologies will compete with low-carbon power generation technologies such as NG-CC with CCS to provide the grid with electricity during times when wind and solar are not producing electricity, we compare them all together within this paper,” researcher Chad Hunter told pv magazine. “This allows for a quick comparison of technologies that have not all been looked at in the same analysis before our paper.”

The techno-economic analysis considered the LDES and flexible power generation technologies in the US Western Interconnection, which is a wide-area synchronous grid stretching from Western Canada to Baja California in Mexico, with an 85% share of renewable energy in the area's electricity mix.

“LDES requires large energy capacities so that a typical rate of charging or discharging can be sustained for days, weeks, or even longer,” the scientists explained. “In this study, flexible power plants and LDES system power generation equipment are sized at 100 MW, in the range of peaking and load-following plant sizes today.”

LDES systems are sized to supply rated power for durations from 12 hours to seven days and the LCOE is calculated for both current and future capital costs.

Through their analysis, the academics found that, for the maximum duration of seven days, NG-CC plants with CCS are the cheapest solution. For the minimum 12-hour threshold, the options with the lowest costs are compressed air storage (CAES), lithium-ion batteries, vanadium redox flow batteries, pumped hydropower storage (PHS), and pumped thermal energy storage (P-TES), which they said is mainly due to their moderate power-related capital costs and high round-trip efficiency.

Popular content

“Batteries will likely play a large role in grid energy storage moving forward, especially if battery prices continue their strong decline as we have seen over the past decade,” Hunter explained. “Shorter-duration battery storage will be complemented by low-cost, longer-duration storage technologies, such as geologic hydrogen storage.”

For more than four days of storage, the least-cost solutions are diabatic compressed air energy storage (D-CAES), NG-CC, NG-CC with CCS, natural gas combustion turbine (NG-CT), and hydrogen storage in salt caverns with re-electrification in heavy-duty vehicle proton exchange membrane (HDV-PEM) fuel cells. They also determined that pumped hydro storage and the HDV-PEM fuel cells with salt cavern storage offer the lowest LCOE for the 12-hour and 120-hour durations, respectively.

“Although hydrogen systems with geologic storage and natural gas with CCS are the least-cost technology options to support high variable renewable energy (VRE) grids at durations beyond 36 h, several challenges are associated with them,” the NREL research team said. “First, neither technology offers the lowest cost for short-duration storage (12 hours), which will likely dominate the storage market until high VRE penetrations are reached; thus, market adoption and learning must be driven by other sectors or use cases, such as using HDV-PEM fuel cells in heavy-duty trucking or deploying CCS for industrial applications.”

The NREL group said that minimizing storage capital is economically convenient at durations longer than approximately 48 hours and that the LCOE is more sensitive to storage energy capacity costs than storage power capacity costs. The team presented its findings in “Techno-economic analysis of long-duration energy storage and flexible power generation technologies to support high-variable renewable energy grids,” which was recently published in Joule.

This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: editors@pv-magazine.com.