Simulating The UK Grid's Transition To Net-Zero - Part 2 (Storage)

Introduction

In the first article of this series, we explored the impact on the electricity generating mix of adding wind and solar generation to the current UK grid. We found that adding wind and solar displaced gas consumption for power generation, but didn't remove the need for gas-fired power plants as these were still needed when the wind was low. We also found that there were diminishing returns to adding more wind and solar with more and more potential output being curtailed because it was being generated at times when it wasn't needed. It was clear that to address this issue some form of energy storage was needed to mop up excess energy and release it again when needed. In this article we explore the addition of energy storage to the UK grid, and in a following article we will look at nuclear power's role in either supporting or supplanting renewables. Given that in the future transport and heating will also need to be made zero-carbon, I will explore this in a fourth article. Once again all code used in the simulation is available on GitHub.

How Can Electrical Energy Be Stored?

One significant advantage of fossil fuels is that they can be readily stored: coal-fired power plants have huge piles of coal waiting to be burned, while gas can be stored in underground caverns, as Liquified Natural Gas or in the gas distribution network and the gas field itself. Of the renewable energy sources, biomass can be stored in wood pellet silos and hydroelectric energy can be stored in reservoirs, but solar and wind are captured by the wind that is blowing and the sun that is shining at any given time.

So what are the options for the storage of excess electricity? The oldest form of storage is pumped-hydro - this involves pumping water uphill from one reservoir to another, and then releasing it back downhill and turning it into electricity. The UK already has pumped-hydro with around 3GW of electrical generation capacity and 30GWh of storage. To be clear, the 3GW relates to the amount of instantaneous power that can be generated by the turbines or consumed by the pumps, whilst 30GWh is the amount of energy that can be stored. So in the UK we can either generate or discharge 3GW of power for 10 hours in theory, although in practice energy losses means that this will be a bit less. For context, though, the annual electricity consumption of the UK is around 300TWh - 10,000 times more than this.

A more recent and rapidly expanding form of storage is grid-scale Li-ion batteries. These are the same batteries as in electric cars, and some of the largest projects underway in the UK come in at around 320MW/640MWh of storage. The benefits of battery storage are that they can be built more quickly than pumped-hydro, and can also be built anywhere whereas pumped hydro needs specific terrain. However, the advantage of pumped hydro is that energy storage capacity can be increased by increasing the size of the reservoir, whereas battery storage and generation capacity are more tightly coupled.

In addition to pumped hydro and batteries, there are a number of additional technologies being worked upon with a variety of different attributes including hydrogen, compressed air, flow batteries, metal-air batteries and gravity storage. A final option is interconnectors to other countries, for example the 1.4GW North Sea Link that connects us to Norway, a country with a huge hydroelectric storage capacity of 80TWh and 35GW of generation capacity. A disadvantage with interconnectors, however, is that whilst they may readily act as a sink for surplus electricity, they may not necessarily be available for electricity import when it is needed if the countries we want to import from also have a shortfall.

Simulation 1: Wind + Solar + Gas + 24GW/500GWh Pumped-Hydro

Our first simulation imagines that we increase our pumped-hydro capacity 10-fold by 2030 to complement 80GW wind, 20GW solar along with an (unrealistic) excess of 50GW gas generation capacity. The pumped-hydro storage acts very simplistically and charges up at any opportunity it gets (ie when wind + solar exceeds demand) and discharges whenever needed (ie wind + solar is less than demand).

The plot below shows how this might work in March 2022 - if the excess electricity exceeds the 24GW pumping capacity or the 500GWh reservoir (represented by the solid blue shading) is full, excess electricity is curtailed. When the wind drops around March 21st (blue line) the energy in the reservoirs is rapidly emptied (blue shading decreases, green line above zero) and when the reservoir is empty gas generation (pink) takes its place. When the wind picks up again around March 30th gas generation can be displaced and the reservoir is refilled (blue shading increases, green line below zero). The question, then, is how much pumped hydro or battery pump/generation capacity (GW) and storage capacity (GWh) is required to minimise the amount we need to fall back on gas generation?

Simulation 2: Wind + Solar + Gas + Storage In Many Configurations

In a similar approach to the first article, we can take the archetypal simulation above, and explore the parameter space of multiple combinations of wind, solar and storage in the time period from 2017 to 2021. In this case we look at systems with 20GW solar; 50GW gas; 40, 80 and 120 GW wind; 0, 100, 500 and 2000GWh of storage capacity; and 12, 24 and 48 GW of pump/generation capacity. The storage configurations are designed to imitate what a predominantly battery or predominantly pumped hydro system would look like: for battery systems the ratio of generation to storage capacity would be 1:2 or 1:4, whereas for pumped-hydro it would be in excess of 1:10. Energy storage has losses but these are not included in this simulation - the aim is not to arrive at exact amounts of capacity needed, only to explore the dynamics and interactions at play.

In the first plot (below) we consider how many GWh of wind, solar and gas derived electricity are generated, with the black dotted line representing total demand over the 5 year period. Any generation in excess of this line is either curtailed or exported. In all scenarios we still generate an excess of electricity that needs to be curtailed, but the amount of gas generation (pink) is reduced the more storage we have (bottom right panels vs top left panels)

The second plot shows how more explicitly how much generation from gas is required (pink), and how much renewable generation is curtailed (red). The amount of gas generation required is lower when storage is available (right hand panels) than when it is not (left hand panels).

In the third plot, we focus on the proportion of gas generation in the mix since this is what we are trying to minimise. Each panel represents the GW capacity of the pumps/turbines whilst the colours represent the size of the reservoir. Broadly speaking the size of the reservoir is more important than the capacity of the pumps/turbines as we get closest to zero penetration of gas with 2000GWh of reservoir and 120GW of wind, where only around 2.5% of generation comes from gas.

In the fourth plot we consider the utilisation factor of the 50GW gas capacity. Although fewer gas derived GWh is better from an emissions perspective, lower utilisation represents poorer economics for the owner of the gas-fired power stations and so would increase the price of these GWh. Ideally we would generate the GWh we need fom as few GW of capacity as possible. Unsurprisingly the gas utilisation looks very similar to the percentage of generation coming from gas.

The final plot explores the maximum gas generation capacity required over the five year period, and in most scenarios it is around 45GW. This corresponds to there being some point over this time period where winter demand is high, wind output is low and the reservoirs are empty. Only when there is 2000 GWh of storage is this substantially reduced, although intriguingly this is most marked when there is the least generation capacity from storage (12GW). This is a somewhat counterintuitive finding, and will be explored in the next section.

Optimising Back Up Capacity

If we look in detail at March 2022 (below), we can see that in the case where there is only 12GW storage discharge capacity, gas generation (pink) and generation from storage (green) work together to get through the wind lull and the reservoir empties slowly (light blue shading) although at the cost of the reservoir also filling more slowly when the wind does come back. Note that storage and gas have been emphasised for clarity.

When there is 48GW storage discharge capacity (below), there is initially no gas generation at all, but the reservoir is emptied quickly until it runs out and then the entire deficit must be made up by gas. However, reservoir is also able to fill back up rapidly and use more of the wind resource when the wind does pick back up.

In reality, storage behaviour would depend on reservoir levels and the forecast for wind, solar and demand in the future. In March 2022 scenario, storage would empty slowly (12GW) and fill up fast (48GW). Price mechanisms based on scarcity of renewables and level of demand would influence behaviour of storage reserves, and also demand. This means that there is scope for gas (or other fuel-based) capacity to be further reduced from the levels modelled here although the amount of generation would remain the same.

Gas capacity and storage are essentially the insurance policy for periods of low renewables output, so it is the expected frequency and duration of these periods that will determine the optimal mix of both. It may be considerably cheaper to maintain some rarely used gas-fired power stations than build extra storage. It may also be cheaper to maintain fewer gas-fired power stations which run more frequently BUT at an increased risk of storage running low in extreme weather or demand events.

Finally, although in these articles we have only considered gas-fired power stations as the fall-back, it is not the only option of dispatchable fuel-based storage. In the UK we have Drax's 4GW of biomass converted coal-fired power stations. By choosing to generate from biomass ahead of gas, and by generating before storage runs out, we can reduce not just the number of GWh of electricity we need to generate from gas, but also the GW capacity we need to retain. The modelling to simulate this is far more complex than that attempted here, but this simulation of the Australian NEM grid is worth looking at from David Osmond.

Adding Extra Solar

The final option to consider is additional generation capacity of a different type. The plot below shows what happens if we have 50GW of solar rather than 20GW.

Now during the daytime there is sufficient generation from solar that the reservoir can be topped up, and this is just enough to make the reservoir last all the way through the entire wind lull without any gas generation at all. It is interesting that even in the UK in March that solar makes a useful contribution to the generation mix in this specific context.

Conclusion

In this analysis we have explored the role of various configurations of storage in a grid dominated by renewables in a very simplistic simulation. We can show that adding storage to our simulated grid helps to drive out gas generation so that it represents at little as 2% of total generation, and that the size of the reservoir is the most important variable as this represents the amount of energy that can be drawn upon during a low renewables generation event.

The importance of GWh over GW in our simulation makes us question the optimum mix of batteries vs pumped-hydro generation - whilst batteries may be useful to provide peak capacity and grid services in the short term, it seems that a renewed focus on pumped-hydro is important and a number of projects have been proposed such as SSE's Coire Glas (1.5GW/30GWh). Arguably these projects should be further optimised to maximise the reservoir capacity.

This still leaves fuel-based generation being required as a generator of last resort, but using storage more smartly could reduce our requirement for little-used capacity, and also focus that usage on biomass rather than gas. Finally, we showed that adding solar can be helpful in providing extra output at times when wind generation was low.

In the next article, we will explore the role of nuclear power. How would nuclear generation support a renewables dominated grid, and what requirements (if any) would a nuclear dominated grid have in terms of support from renewable generation, storage and gas-fired generation?