Given the centrality of the power sector to the U.S. economy, it’s no surprise that debates rage around how to accomplish a clean, reliable, cost-effective grid. Over the years, numerous studies have been published that evaluate this topic, particularly trying to answer the question of what resources are necessary to maintain reliability under various policy or clean energy scenarios, whether these are set at the state or utility level. At Form Energy, we live and breathe energy storage, so we’re naturally interested in what these new studies mean for batteries and beyond. In this piece we zoom in on three such studies, which analyze how various regions in the U.S. can meet their electricity needs with a variety of clean resources.

The most geographically focused of these, the LA100 study, conducted by the National Renewable Energy Laboratory explored several pathways for the Los Angeles Department of Water and Power (LADWP) to meet its electricity needs with reliable power by 2045, in line with LADWP’s stated portfolio objectives. A second study (Long et al.), from leaders in academia, consulting, and nonprofits, focuses on California, while a third study (Sepulveda et al.), from MIT and Princeton academics, explores regions like Texas and the Northeast.

The three studies echo a common conclusion in the growing literature on power grids: while solar, wind, and lithium-ion batteries will provide valuable energy and grid services in the future, the presence of firm resources provides significant cost savings and reliability benefits over a renewables and lithium-ion only approach.

What follows is our summary of the implications of these studies for the energy storage industry, complemented by our original analysis.

Spoiler Alert: Cost and Duration of Storage Matter. A lot. 

While the various studies take different approaches to modeling storage, they all conclude that ultra-low cost storage can decrease the costs of grids substantially and that the capability to generate over multiple days of adverse weather is critical for reliable power.

Sepulveda et al. conclude that storage technologies with energy capacity capex costs less than $20 per kilowatt-hour can save billions in electricity costs relative to systems with only renewables, lithium-ion, and carbon capture, nuclear, or hydrogen. Furthermore, they conclude that storage technologies “exceeding 100 hours” in duration play the biggest role in reducing power system costs.

LA100 finds that hydrogen paired with onsite storage and converted to electricity in a combustion turbine or fuel cell (the only long duration storage technologies modeled) could be a key pillar in achieving LADWP’s targets. This is true, even at the high costs that LA100 modeled – $3,200 per kilowatt to $5,300 per kilowatt depending on year and technology, or three to five times the cost of today’s combined cycle power plants. The LA100 study doesn’t explicitly define the duration of the hydrogen storage technologies modeled. Rather, it states that the technologies have “many days of capacity” and finds that grid reliability is maintained by discharging for “extended periods over multiple sequential days [to] ensure load balancing on consecutive days or weeks with low wind and solar resource availability.” This value is especially apparent in constrained grid areas like the Los Angeles basin in California.

Long et al. don’t explicitly model long duration storage in their core cases. However, they do model “zero carbon fuels” that include hydrogen, at costs ranging from $15 per MMBtu to $50 per MMBtu. While Long et al. don’t model it as such, zero carbon fuels like hydrogen that are produced with electricity are a form of long duration storage in that the fuel must be created and stored for later use. In turn, they find that long duration hydrogen storage is necessary to cost effectively maintain reliability “when the sun doesn’t shine for many days.”

While all of these exercises point to the value of storage with multi-day discharge capabilities, all studies also point to the fact that a technology inclusive, portfolio approach delivers the least cost, most reliable electric system, a finding we have confirmed many times over in our own modeling.

Can storage be “firm”?

While these studies and others point to the value of firm technologies, few provide a clear definition of “firm.” The inclusion of hydrogen as a “firm” option in these studies adds further confusion, as hydrogen requires storage and other supporting infrastructure and its availability is subject to weather vagaries. Aligning on this definition can help provide clarity to policy makers, regulators, utilities, and entrepreneurs.

In practice, all resources are subject to availability conditions. As we learned from the February 2021 storms in Texas that brought down roughly one-third of the state’s thermal generation, no technology is perfect. Ultimately, firmness is defined by the relevant regional weather and demand conditions, and firm resources must be able to consistently meet a given demand under a wide variety of those conditions.

To understand what those conditions could look like in one of the focus regions of these recent studies, we analyzed 35 years of intermittent generator profiles from the solar and wind generation datasets used in the California Public Utilities Commission’s (CPUC) Integrated Resource Plan to determine what types of resources might be necessary to provide firm power in California.

Our findings show that periods of low intermittent generation¹ lasting between 50 and 100 hours occur fairly often (roughly once every two years) and that periods lasting from 100 to 150 hours occur roughly once every ten years. In one of the most severe of these events, which occurred in December 2010 and is shown in Figure 1 below, renewable output was well below expected levels for six days.

Figure 1: December 2010 low-renewables event in California

Even under more moderate definitions of low generation (where “moderate” means extended generation shortfalls that are within the Planning Reserve Margin of 15%), we did not find a single event lasting longer than 150 hours. In other words, in California, storage systems that can discharge at full power for at least 100 hours could provide clean firm power in all but a 1-in-10 year weather event, during which 150-hour systems would be sufficient.

This echoes findings from a February 2021 study of the Northeast U.S. from ISO-NE and DNV GL, which found that wind lulls lasting three days occur annually, while five day lulls are less common. While our analysis focuses on California, the underlying method is broadly applicable and can be applied for regionally-specific, data-driven definitions of firm power.

In summary, these recent papers and an examination of wind and solar energy availability in California point to the need for a new class of storage – multi-day storage – that can provide reliable power to grids across multiple consecutive days of low renewable energy output.

Scaling firm, multi-day storage

Researchers are increasingly aligned on the need for firm technologies. Recent work has clarified that, for storage to be truly firm, it will need to reach dramatically lower costs and longer durations than today’s short duration technologies. Data-driven and model-based analyses point to the need to provide dispatchable energy for more than 100 hours to be truly firm, and at a price point which enables that, which starts at ~$20/kWh. The benefits to society of scaling multi-day storage would be enormous.

To meet the grid’s coming challenges, low cost, reliable, and clean resources will need to be built out at a record pace; transmission will need to be upgraded; pipeline networks will need to be retrofit or built from scratch; and low cost storage capable of providing multiple days of firm power will need to be scaled.

There has never been a better time for visionary infrastructure, innovation, or decarbonization plans to meet the challenges of the energy sector.

¹ We defined low intermittent generation as having a 3-day moving average hourly output that was at least 25% below average. This amount is roughly consistent with two standard deviations below average.