We need to decarbonise electricity generation.
Electricity generation worldwide accounted for 12.3 Gt CO2e in 2020, or around a third of total emissions. One of the key ways to reduce this is to increase deployment of clean energy from renewables. The IEA’s net zero 2050 projection, which keeps warming below 1.5℃, requires 61% of electricity from renewables by 2030 and 88% by 2050 and in the UK the government has committed to decarbonising the grid entirely by 2035.
Fortunately, wind and solar are now the cheapest form of power available: in 2021 nearly two-thirds (163 GW) of newly installed renewables had lower costs than the cheapest coal-fired option. This fall in cost, coupled with the global energy crisis, is leading to widespread deployment of wind and solar; we are set to deploy as much new renewable power in the next five years as we have in the last twenty.
However, wind and solar are intermittent and so not dispatchable, that is, they cannot be ramped up on demand. When coupled with an increase in overall electricity demand, due to population increases, increases in living standards, and the electrification of additional sectors of the economy—including battery electric vehicles for transport and electric heat pumps for home heating—the result is some big upcoming challenges in managing the grid.
The need for grid flexibility
On the grid it is important to balance electricity supply and demand—what goes in must come out (and vice versa). To achieve this grid operators look for generation options that can meet 3 different types of demand profile:
Baseload (or firm) generation refers to the minimum energy demand that is required for a given day. Nuclear is the prototypical base load power – it is difficult to ramp up and down quickly, but can provide a steady, reliable stream of power.
Load levelling describes the predictable intraday fluctuations in supply and demand, and is traditionally provided by gas- and coal-fired power stations.
Peak shaving refers to the instantaneous adjustments in supply to match demand. It requires relatively rapid response times (from seconds to a few minutes) in order to curtail production in the case of too much supply or to ramp up supply in case of a shortfall. This requires highly dispatchable generators, and is currently provided for by specialised (relatively expensive) peaker plants.
Renewables struggle to participate in all of these situations. Wind power can provide fairly reliable base load in a daily cycle, but is subject to seasonal fluctuations and longer duration drop outs—this is referred to as ‘Dunkelflaute’ in German. The ‘California duck curve’ above neatly illustrates the challenge of matching the peak in supply of solar–at midday when the sun is shining–with the predictable peak in demand later in the day, around 6pm. And renewables intermittency is doubly challenging for peak-shaving, causing them to be curtailed (i.e. turned off) some of the time and, of course, not being able to rapidly ramp on-demand.
In addition to ensuring supply, the transmission and distribution of electricity on the grid requires specific attributes that are provided (largely by generators) as ancillary services. These include frequency regulation, inertia and voltage control. Whilst the technical details are beyond the scope of this post, it is important to note that increased deployment of wind and solar negatively impacts the stability of the grid, and so alternatives that can provide these services will have market opportunities beyond wholesale energy markets.
So how can storage contribute?
As you can see, one of the key aspects of the grid today is being able to generate electricity at (approximately) the same time that it is needed. Using energy storage enables us to decouple this dependency. For many years, the main form of storage on the grid has been pumped-storage hydropower (PSH). When we have had excess electricity we’ve used it to pump water uphill so that later we can let it run down hill and drive a turbine to create electricity. For the relatively limited time-shifting needs of the old fossil-fuel powered grid, this was sufficient.
But now that we have intermittent renewables, there is an opportunity for a significant amount of additional storage to be added. Areas that are particularly interesting are:
Very large capacity and long duration storage to enable reliable, renewable base load in the face of Dunkelflaute.
Medium duration (4-12 hours) storage to shift predictable peak solar generation energy toward predictable peak usage later in the day
Short duration storage to help manage the unexpected peaks and troughs, allowing renewables not to be curtailed and preserving the responsive grid we all love
As mentioned, PSH is the only storage technology deployed at scale but the deployment of grid-scale lithium-ion batteries is growing extremely fast. To date, these batteries have generated most of their revenue providing (more lucrative) grid stability services (frequency response), but are increasingly being used to provide load levelling too, as seen for California above.
Other grid flexibility strategies that will also play a role include: overbuilding of wind and solar, demand-side response, interconnectors and the use of clean dispatchable resources such as nuclear, geothermal and hydroelectric power. Demand-side response and interconnects will undoubtedly play an important role, but won’t be sufficient for very long duration and large capacity drops in production (not to mention the geopolitical challenges on over relying on interconnects).
The success of storage vs overbuilding vs clean base load will be a question of cost. Nuclear is clean and reliable, but it is expensive and has long lead times (and in places struggles with poor public perception). Geothermal energy, on the other hand, is currently expensive and geographically limited–although technical innovations in this space could make it
cheaper and more widely adopted. The cost (levelised cost of storage or LCOS) of grid-scale lithium-ion batteries (including the cost of charging with solar) is already cost competitive with gas peaker plants (around $150/MWh): so batteries are as cheap and as performant as peakers for intraday shifting and peak shaving. However, longer term storage (or indeed clean base load) will need to be cheaper: for the longer duration storage, the target is to be below $50/MWh.
Regulation and market design will be required to stimulate the market for storage
The need for alternatives to support renewable penetration is clear. It is hard to predict the market size at this early stage, but this McKinsey study suggests that up to 2.5 TW and 140 TWh of storage will be required. The challenge of energy storage is framed here as a competition between firm base load generation and storage, and–crucially–which can be deployed more cheaply. Really, it is likely that both will be required and play an important role.
New regulations and market design will be required to encourage investment into storage assets. The policy support is there: there are at least 59 policies globally in support of grid-scale storage deployment, according to the IEA, with the majority having been implemented since 2020. However, the current structures of wholesale and capacity markets do not guarantee income for storage assets; it is likely (and necessary) that storage-specific markets will be designed as renewables penetration increases.
In the next blogpost, we will look at some of the candidate technologies for energy storage and explore opportunities for innovation and investment.
Special thanks to Sebastian Blake (Octopus Energy) and Dr Iris Ten Have (Extantia) for useful discussions regarding energy markets and technologies.