TL;DR

While battery costs have plummeted, truly long-term energy storage beyond daily cycling requires significantly cheaper solutions than current lithium-ion technology can likely deliver. We do not yet have technology that can service storage durations on the order of week or longer. Emerging technologies show promise but face their own challenges.

Using LCOS to determine required installed costs

The plummeting cost of batteries has been incredible, but a crucial question remains: how much cheaper do they need to become to truly power a renewable energy future? Battery viability is a function of both expected returns and costs. In this post we will focus on the cost portion assuming that a certain cost implies desired economic viability. Furthermore, we will focus on the useful metric Levelized Cost of Storage (LCOS).

LCOS gives us the per unit cost of storage amortized over its lifetime, considering the time value of money and cycling frequency. In simple terms it is the discounted cost over its lifetime divided by the discounted energy stored over its lifetime. In doing so, we can compare different technologies even if they have differing cost structures, lifetimes, degradations and so on. See this post for a more in-depth explanation of LOCS.

For this post, we define the average cycling frequency as how often a battery is fully charged and discharged per year. If a battery is discharged 52 times a year, we say it is cycled on average weekly, even if it was in fact cycled 30 times in one month and 22 times spread out over the remaining year. By talking about daily or weekly average cycling frequencies, it is easier to conceptualize. As more batteries enter the market, the frequency with which they can discharge reduces as competition increases.

We will be using LCOS to give us a sense for what installed costs (numerator) we need to achieve to service different cycling frequencies (denominator). As the market for one and two-hour duration battery systems become saturated in several markets, batteries are being asked to increase their storage capacity to service longer and longer durations. In a future post we will dive into what frequencies we need to service for deep decarbonization, but here we stick to conceptually useful frequencies.

Let’s jump in! Given a desired LCOS and cycling frequency, we can rearrange the LCOS formula to return the necessary cost structure. While the full LCOS considers the Net Present Value (NPV) of all costs, for this simplified analysis, we're focusing on the upfront installed cost, assuming negligible ongoing operating expenses. We performed this simple analysis for daily, weekly, monthly, and annual average cycling frequencies and a target LCOS of $0.05/kWh and a discount rate of 8% (reasonable opportunity cost/cost of capital for average renewable energy investor). This target LCOS is based on DOE’s Long Duration Storage Shot intended to support increased renewable penetration. It further serves as a reasonable cost for performing arbitrage in energy markets.

| Average Cycling Frequency | Required Installed Cost |
|------------------|-----------------|----------------|
| Daily (365x/yr) | $144 |
| Weekly (52x/yr) | $21 |
| Monthly (12x/yr) | $5 |
| Seasonal (4x/yr) | $2 |
| Annual (1x/yr) | <$1 |

Table 1: The required installed cost for storage to achieve a LCOS of $0.05/kWh assuming no losses, no degradation and different average cycling frequencies across a year.

In Table 1 we see that with daily cycling (roughly what battery system do today) we need an installed cost of $144, which is something that we are close to hitting today (as low as $200/kWh in the US). For once-a-week cycling we need to lower our installed costs by a factor of 7 or $21/kWh. This is currently well below the LFP pack prices ($115/kWh according to Bloomberg) and certainly below the installed costs, which means we need to improve both on the LFP manufacturing costs as well as the installation costs. As we move further to monthly, we need to do about a factor of 4 better, down to about $5/kWh which is close to the material cost of LFP batteries (~$8/kWh). This starts to seem unreasonable for traditional LFP batteries.

How cheap can we get using learning curves?

Another way to look at the feasibility of these prices is by using learning curves, since these are often referenced for battery and solar costs decreases. Looking at historical data from Bloomberg, the learning rate for lithium batteries is roughly 18% per doubling. If we continue this learning trajectory, how far do we get?

Let’s start with global electricity generation at 27,000TWh in 2023. This would require us to increase our current capacity by a factor of one thousand (which will most likely require a much larger factor in nameplate capacity*). This means we need to double about 10 (2^10=1024) times. With a learning curve at 18% we would experience an 86% decrease in cost, putting us at $28/kWh. As Table 1 shows, we need to hit $21/kWh for weekly cycling to be economically viable at our target LCOS. This suggests that even with significant continued cost reductions, lithium-ion batteries may still fall short for this duration. Of course, storing all global electricity generation is unrealistic, but this is just to give us an idea of what we could hit if we continue our learning curve indefinitely.

However, we have a lot more energy needs than electricity, so if we stretch to our total primary energy consumption (ignoring for a minute energy efficiency, heat pump factors, etc.) we would need 180,000TWh. This would require about 16 doublings to reach which would result in a 96% decrease in price (about the same as 2000 to today!) and put us at $8/kWh. Interestingly, this is roughly the material cost of LFP batteries today. It is thus from a physical (material cost) and theoretical (learning curve) perspective going to be difficult to get lower than this. Fortunately, this would allow us to service up to bi-monthly frequency, but still put seasonal storage out of reach.

The point here is that even in the most optimistic cases it is extremely unlikely that electrochemical batteries will be our storage mechanism for long term storage longer than a week.

What candidate technologies do we have?

Table 2: Summary of different long-term storage technologies. All of these are attempting to get below $100/kWh and ideally down to $10/kWh.

Let’s see how other technologies compare. In Table 2 below we summarize different candidates for low cost long-term storage. Pumped hydro currently commands the largest market share and has a low end cost estimate of $100/kWh. This puts it below current LFP project costs, but still above the cost level we need for weekly average cycling frequency. Unfortunately this technology is mature (low expectation for future cost reductions) and geographically constrained for further deployment.

Compressed air and hydrogen have received a lot of attention recently given their ability to store vast amounts of energy at low cost. Unfortunately, even these technologies still remain too expensive for lower cycling frequency although we may see larger cost reductions relative to pumped hydro for these relatively newer technologies. It should be noted that these technologies have poor round-trip efficiencies, so it is important to consider this in addition to the installed cost.

Iron-air batteries like the ones being developed by Form Energy have received not only a lot of attention but also funding based on their incredible claims of $20/kWh. If they can indeed reach these low costs, they could be a viable candidate for week long storage. Similar to CAES and hydrogen however, these batteries have poor round-trip efficiency which may still preclude them from offering economically viable storage for weekly cycling frequencies.

Perhaps the most promising is thermal energy storage which is already reaching costs that would make weekly cycling economically viable. However, it is important to note that the thermal storage we are considering here is only capable of storing thermal energy, which means we still need technology that can provide electricity storage at longer durations.

The role of transmission

In addition to battery storage, transmission can play a huge role in stabilizing the grid and moving energy from oversupply to undersupply locations. This will require massive infrastructure, but could for example provide excess solar from California during the day to serve the evening peak on the East Coast. Similar to many of the battery technologies listed above, transmission is relatively mature in terms of learning curves and the costs remain prohibitively high ($M per mile) for the large scale transmission infrastructure that we would need to avoid the need for storage. However, strategic placement of storage and transmission may help us reduce the overall capacity of either, lowering our total costs and providing complementary types of grid stability.

We have our work cut out for us

Despite the ambitious targets of companies like Form Energy, we are still quite far from economically serving long-term energy storage. There is a lot of necessary work to do on storage if we expect to increase the penetration of renewable energy. The DOE Long Duration Storage Shot indicates that we need to reduce our current lowest storage costs by a factor of 10 to achieve desirable levels of renewable penetration, enough of a reduction to be considered a “moonshot.” As short duration energy markets are saturating, we need to urgently consider what comes next after LFP. We have a variety of potential technologies, but we aren’t there yet.