Long term electricity storage is wishful thinking
Authors: Yogesh Upadhyaya and Manish Agrawal
My co-author and I have long talked about the unreliability of wind and solar power plants and the costs it imposes on grids. The unreliability requires wind and solar power plants (collectively referred to as Variable Renewable Energy or VRE) to be fully backed by conventional power plants. This is especially so for the costs imposed by seasonality. That is, the times when the wind does not blow for days or the Sun does not shine for weeks. Many people talk vaguely about how long term storage costs would fall ‘some time in the future’ and solve these problems. This is just wishful thinking. Let me explain.
Let us classify the unreliability of wind and solar power plants in three types. The first type is very short term. The output of solar panels can drop significantly because of a passing cloud. Similarly, the output of wind turbines can drop sharply if wind stops blowing. This intermittency can impose significant reliability costs on the grid. However, the volume of storage required for this kind of intermittency is low. Of the order of a few seconds to a few minutes.
The second kind of unreliability comes from the so-called Duck Curve. In many grids, peak demand may occur in the evening when the Sun has stopped shining. In fact, the peak occurs precisely because the sun has stopped shining and people go home and switch on the lights and cooling devices. This means that the peak demand comes at the same time that solar panels stop generating. This imposes a demand on the conventional power plants to increase generation very quickly. As the share of solar capacity increases in a grid, the required ramp up becomes steeper and steeper and coal plants especially cannot keep up. Storage can help. This storage has to be of the order of a few hours. This is already a problem in India. The Indian electricity operator recently warned that in the worst case, there is one in three chance that supply falls short of demand in the coming months. The Central Electricity Authority is now ‘urgently’ asking solar power plants to deploy storage solutions. The next time someone tells you how cheap solar panels become, remind them to add the cost of at least a few hours of storage in their calculations.
A few hours storage is much more costly than that for a few minutes but either of these costs is nothing compared to the storage cost that is required to deal with seasonality. Let us take an extremely simplified example to understand this.
Assume that you have a grid with a flat demand of 100 MW. The grid is supplied by 80 MW of thermal capacity and 25 MW of wind capacity. The storage capacity that you would need for different kinds of unreliability is calculated below.
To back up no wind period of one hour you would need storage capacity of
20 X 1 = 20 MWH.
To back up no wind period of four hours you would need storage capacity of
20 X 4 = 80 MWH
To back up no wind period of five days you would need storage capacity of
20 X 24 x 5 = 2,400 MWH.
Do note that a period of 5 days with no or low wind in real grids is not at all unlikely. In Germany, the generation from a wind capacity of 70 GW can fall as low as 0.2 GW for days on end in the Dunkelflaute days. Even in India, wind can vary quite a lot. Here is the VRE generation for the month of May 2024.
As you can see, the generation between 10 and 16 May was ⅙ of the of the peak! Remember, this is the aggregate generation. In specific regions and plants, the variation would be much higher. Variation in solar plants could be worse. India has four months of monsoons. In my native city of Mumbai, one may not see sun for 15 days at a stretch in those months. Other places in India are not as rainy but 5 cloudy days in a row is not unusual. And the monsoon months are followed by winter which reliably has lower sunlight hours. If you add the impact of dust in the atmosphere, the fall in capacity of solar power plants becomes worse.
As our example showed, the storage capacity needed to compensate for long periods goes up sharply. A 120 X increase from 20 MWH to 2400 MWH! However, this is not all. The big problem of long term storage is that it will be used very less. This becomes obvious when you look at the per unit costs.
Let us assume that the cost of storage of one hour is Rs. C / year, for 4 hours Rs. 4 C / year and 5 days Rs. 120C / year. Note that these would be mainly capital charges such as debt servicing as most storage options are very capital intensive.
The short term storage may be used multiple times a day. If it is used three times a day, the number of units stored and released in a day would be 3 x 20 MWH = 60 MWH. In a year, it would be used 60 X 365 = 21900 MWH.
The medium term storage would be used once a day. So in a year it would be used for 80 x 365 = 29200 MWH.
It is not clear how many times the long term storage would be used. Let us assume it is used 5 times. That would be 2400 x 5 = 12,000.
So the long term capacity would not only be much more expensive than the short and medium term capacity, it would also be used over fewer units.
The capacity cost per unit for the short term storage would be C/21900
The capacity cost per unit for the medium term storage would be 4C/29200
Whereas, the capacity cost for the long term storage would be 120C/ 12000 which is 250 times that of the short term capacity cost per unit!
More formally, storage is a capital intensive asset. The main costs for such assets are financing costs. Financing costs per unit of use are very sensitive to how many times the asset is used. The lesser the asset is used, the more are the per unit costs. As a rough analogy, think of a 1,000 Crore airport financed by passenger charges. If 10,000 people use it every day, the passenger charges would be much lower than if the airport was used only 5 days a year and hence used by 50,0000 people in the whole year.
Note, that this challenge for electricity storage is not related to technology. When people say ‘when long term storage becomes cheaper’ they are vaguely thinking of a technology solution. The problem is not technical. It is arising from the nature of VRE. VRE cannot be relied on and hence needs backup. If the backup is going to be used only for a few days a year, then the per unit cost of that backup will be astronomical.
Now, I know that I have made many simplifying assumptions. For instance, the daily demand in my example is flat, whereas in a real grid it varies over the twenty four hours. But I don’t see why that would matter. The seasonal requirement for storage would be in addition to that for daily variations.
Proponents of VRE would argue that some of the demand can be persuaded to shift to the time that sun is shining and wind is blowing. I would say that experience is showing how little demand can be shifted especially across seasons. It is not that you would charge your car less in monsoons because the sun has been covered by clouds for a week. Sure some demand can be shifted. For example, power to agricultural pumps may be given only during bright days. However, there is a limit to how much load can be shifted.
Another simplifying assumption that I have made is that there are only two types of generation sources — thermal and VRE. Proponents of VRE are fond of arguing that wind and solar may not be correlated. They would point to simulations where a combination of different sources of VRE ‘works’. My question would be if the model makers would be willing to guarantee their results? Because girds have to. In any developed society, the grid has to guarantee a very high availability.
We can point out many other simplifying assumptions. However, that would be missing the point. The cost of long term storage is orders of magnitude higher than that of short term storage (which itself should be added to the VRE generation cost). If different assumptions reduce this differential from 250x to 120x in a model, would it really matter?
The load in a grid varies. This variation is over the day as well as over the year. The suppliers of electricity need to be able to respond to this variation. They have to do this reliably as well as cheaply. Coal power plants supply nearly two thirds of electricity in India because they are reliable and cheap. Incidentally, coal power plants have always had a very cheap storage of energy. Here is a picture.
That’s right. A thermal plant stores energy as coal and generates electricity only when it is needed by the grid.
Wishful thinking is seductive. One could argue that it has been at the heart of all technical progress. For example, two brothers had a dream of humans flying and they worked hard at making it a reality. Today we take flying for granted. However, wishful thinking that ignores reality is not very useful.
We may all wish for an electricity source that does not pollute or release Carbon in the atmosphere, is cheap and provides electricity reliably. But we have to do more than just wishing. And till real solutions emerge, we have to live with the reality.
This story is part of a series called ‘Electricity deep dive’.
You can follow AskHow India (@AskHowIndia) or me (@YogeshUpadh) on twitter or LinkedIn or on https://t.me/YogeshUpadhyaya
Follow Manish Agarwal (@manishbgagarwal) or find him on LinkedIn
