How much storage do we need and how long-term is long-term?

TL;DR

For a fully decarbonized German electricity grid, we would need about 191TWh of storage capacity with most of it seasonal and annual storage by capacity (92%). However, weekly storage delivers the largest share of energy (38%) given the higher frequency of cycling. Even massive overbuilding won’t eliminate the need for long-term storage and comes with prohibitive costs. Thermal storage is a realistic solution to the storage problem but is still in the early stages of adoption and has some serious hurdles.

Storage needs to go long-term

In many countries we are seeing increasing deployment of renewable energy sources but also more issues with intermittent renewables from power shortages to extensive curtailing. To address the issue of intermittency we need to look to more transmission and storage as the answer, in this post we focus on storage. Let’s explore the storage challenge heuristically, what is the size and scope of the problem without technical concerns?

To understand the scale, let’s use a simplified model for the German grid (because they have extensive open data) based on the following assumptions:

• Perfect transmission across Germany (no congestion)
• 100% efficient batteries with no losses
• 2024 Germany electric grid hourly energy mix and consumption
• To model a decarbonized grid, we remove all supply currently produced by fossil fuel plants. For the remaining “clean” grid, we then scale the remaining annual supply to match the annual demand.

If we now model a hypothetical battery that charges and discharges to fill in the gaps of renewable intermittency and variable demand, we find that we need a total storage capacity of 191TWh to bridge all the intermittency and demand gaps.

Limitations

A few items to note: (a) these are some very optimistic assumptions, and in practice we will need much more capacity, (b) these estimates are based on the current grid demand and supply, so if we continue to decarbonize industries like transportation and HVAC, we will see greater needs for electricity and storage, (c) this doesn’t include possible reductions in energy use with energy efficiency. All these items will dramatically change the numerical results, but the general theme of requiring extensive storage and much of it being long-term storage will persist.

Capacity vs Work – why cycling frequency matters

In Figure 1 below we break down the capacity by cycling frequency. Cycling frequency definitions are based on how often that storage unit was cycled in the modeled year. Daily implies 365+ cycles, weekly 52+ cycles, and so on until 1 cycle is annual. Based on capacity (TWh), storage cycled infrequently - annual or seasonally (1-4 cycles per year) – makes up the dominate share (~92%). This means most of what we build needs to be incredibly cheap (<$3/kWh installed).

Cycling Frequency Capacity Shares of Total Storage Capacity

Figure 1: The shares of the total 191TWh storage capacity that each storage duration (daily, weekly, etc.) represents.

However, when we consider the work done (ie. energy delivered over the year), then weekly cycling is the workhorse for the grid. In Figure 2 we show the share of total energy delivered by storage at each cycling frequency (a weekly storage unit delivers 52+ times more energy than an annual unit per year). The picture changes dramatically for work, weekly being the true workhorse at 38% of all energy being cycled at a weekly frequency. Annual and seasonal storage combined only deliver about a third of the energy.

Cycling Frequency Shares Based On Energy Delivered

Figure 2: The shares of the energy delivered that each storage duration (daily, weekly, etc.) represents.

This is significant because storage cycled more frequently (like weekly) can potentially command higher revenues and doesn't need to be as ultra-low cost per kWh of capacity as seasonal storage to be economically viable. Let’s further break down weekly into shares by day frequencies (Figure 2). We now see that 2-day and 3-day cycling frequencies dominate here and are each responsible for roughly one third of the weekly energy delivered with the remaining 4-7 day durations delivering the last third. Figure 3 shows that even within the broader category of storage cycled weekly or less often, a significant portion of the energy delivered comes from units cycling every 2-3 days. This highlights that while seasonal capacity dominates, the energy workhorse includes durations shorter than truly 'seasonal'.

Breakdown of Weekly Cycling Frequency Capacity Shares

Figure 3: Breakdown of the shares of the energy delivered that each multi-day storage duration represents. Cumulatively these make up the 38% weekly share of the total energy delivered.

How feasibly can we address these storage needs?

In Table 1 we tabulate the required installed prices for storage to achieve a $0.05/kWh Levelized Cost of Storage (LCOS, more info here). This target was set by the DOE Long Duration Storage Shot and we believe this LCOS target to be economically competitive for storage systems operating in electricity markets.

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.

Two (2) and three (3)-day cycling frequencies require an installed cost of roughly $48-72/kWh to yield a competitive $0.05/kWh LCOS. Lithium Iron Phosphate (LFP, current industry standard) battery pack prices in 2024 are approaching this range of $48-72/kWh. However, utility-scale total system installed costs (including inverter, balance of plant, etc.) are still higher around $200/kWh in 2024 for short duration systems. For longer-term storage installations, we can consider the marginal cost of adding duration to these systems. Since battery pack prices represent a large portion of these marginal costs and are approaching the target range of $48-72/kWh, this suggests the marginal cost of increasing storage duration with LFP is getting close to what’s needed for multi-day cycling.

However, as we cycle at a weekly interval or less frequently, we need to achieve installed prices of $21/kWh. LFP battery pack prices would need to come down by a factor of at least 4 to reach this, which is going to be challenging. In a different post we investigated how feasible this is.

Won’t overbuilding renewables solve this?

Some propose building vastly more solar and wind capacity than needed (“overbuilding”) to minimize storage requirements. Let’s test this and multiply our solar and wind supply values from our model by 5, keeping the baseline (hydro, biomass, hydro) the same. We see that our total storage needs drop dramatically from 191TWh to 20.5TWh. Interestingly, the energy-based shares stay roughly the same with weekly accounting for about 37% of energy delivered from storage. This indicates the magnitude of the problem is reduced, but the nature of the problem (ie. relative need for different durations of energy delivered) remains.

Let’s look at some of the consequences of overbuilding. Assuming base renewable costs of $0.03-$0.10/kWh, a 5x overbuild increases the renewable generation portion of the LCOE to $0.15-$0.50/kWh – a massive and likely unacceptable increase in overall electricity costs. This is of course a simplified picture but illustrates the serious consequences of overbuilding on costs. We also have a post on overbuilding and why it won’t happen.

Is this a universal pattern? CAISO as another example

To see if this pattern holds elsewhere, we repeated the same analysis above for the CAISO grid (California) in a sunnier climate. The total storage needed is 114TWh and again, the share of total storage capacity is dominated by seasonal (40%) and annual (59%) needs. As with Germany, despite seasonal/annual storage dominating capacity, weekly cycling provides the largest share of energy delivered at 56%. The fundamental challenges remain: Cost-effective storage across various durations is necessary for full decarbonization with multi-day and weekly cycling being critical in the near term.

Where might this not hold?

There are a few obvious regions where this analysis doesn’t hold true and a different set of challenges exist. For example, this analysis assumes a strong reliance on solar and wind resources. In countries like Norway, hydro power dominates the energy mix and reduces the intermittency issue that we modeled.

This analysis also doesn’t hold for countries in Southeast Asia such as Singapore or the Philippines where there is little summer-winter supply and demand disparity. In these countries we expect storage capacities to be much lower and cycling frequencies to be much higher. This analysis also clearly doesn’t hold for countries that are not actively decarbonizing such as the middle east, where storage has little to no value.

What other options do we have?

There are a variety of technologies currently being developed and deployed for longer term storage including iron-air, pumped hydro, compressed air, and gravity-based systems. This post explores these more. In short, these technologies still fall short of the price target for weekly cycling and, as of yet, are far from achieving the price targets for seasonal and annual storage.

Thermal energy is an option, but only addresses a subset of the energy problem

Perhaps we are too focused on storing and delivering electricity. Much of the energy storage challenge, especially seasonal, involves shifting energy captured during the sunny summer months to cover heating needs during dark, cold winters. This is fundamentally a thermal problem, not an electricity problem.

Technologies like large-scale tank or pit thermal storage are remarkably simple and cheap. Recent utility-scale projects in Europe have demonstrated costs around EUR30/m3 (approx. $35/m3). Assuming a temperature difference of 20C (storing heat at 40C and using it down to 20C), 1 cubic meter stores about 23 kWh of sensible thermal energy based on water density, specific heat, and the temperature difference. This puts low cost thermal storage at $1.40 per kWh thermal energy (kWhT).

To compare this to a LCOS per electrical unit, we need to consider that heat pumps can convert electricity to thermal energy at an efficiency greater than 100%. We will consider an average winter coefficient of performance (COP) of 2-3 (since COP varies with temperature), meaning we move 2-3 units of thermal energy per unit electricity. To get the equivalent cost in terms of electricity we can multiply the thermal storage LCOS by the COP:

\[ \text{\textdollar}1.40/kWh_{T} * (2-3 COP) = \text{\textdollar}2.80-4.20/kWh_{e-equiv} \]

This calculated cost of $2.80-$4.20 per kWhe-equivalent is astonishingly low, particularly when compared to our <$3/kWhe installed cost target for economic seasonal electrical storage. This analysis doesn’t consider thermal losses which happen in real world systems. However, we also don’t consider the COP efficiency boost when charging the thermal storage during warm summer periods. We believe that while we have system inefficiencies, the two factors mentioned will yield real-world results that are still reasonably close to our modeled estimates above.

Critically, this solution only addresses the thermal energy need. It cannot provide electricity to the grid during periods of low renewable generation and high electricity needs.

Unlocking thermal: The heat pump activator

The incredibly low cost of storage makes large-scale thermal storage a compelling solution. The challenge? Integration. First, effectively using cheap thermal storage for heating and cooling requires an accompanying heat pump system. While the thermal storage itself is cheap, the heat pump system represents a significant upfront capital cost which can hinder overall system adoption. Second, large-scale thermal storage uses water as a storage medium which requires the ability to push and pull heat in and out of water. Common mini-split and air-to-air systems aren’t capable of working with water. For this we need air-to-water and water-to-water systems.

Fortunately, there is a growing interest in monobloc (also hydronic) systems natively compatible with thermal energy storage. Promoting the adoption of air-to-water and water-to-water systems in place of mini-splits and air-to-air systems will be key for unlocking the potential of cheap, large-scale thermal storage.

Near term needs are within grasp, but we have a way to go for true long-term storage

While utility-scale LFP batteries still face cost hurdles to fully cover the weekly storage need, their decreasing costs are bringing them within reach for shorter multi-day durations (e.g., 1-3 days), which, as shown in Figure 2, accounts for a substantial share of the energy needed from longer-term storage. Emerging long-duration technology (like iron-air, pumped hydro, etc.) holds promise for covering the longer end of weekly durations and seasonal needs, although they are not yet cost-competitive for these applications. With thermal storage we also have a powerful technology to reduce the magnitude of the problem at compelling low cost.

It is worth considering, however, that we may never build out this storage, as deploying peaker plants (likely fossil fuel) may remain more economical in the long run, especially for the very long-term storage needs. Building infrastructure is incremental and unless we bring storage costs down sufficiently, we may never achieve full decarbonization. But perhaps 95% decarbonization will be sufficient after all.