The problem is the duration.
I agree that pumped storage is the best storage solution we have available, but none of the available storage technologies are typically sized for providing power all night, they are typically just a couple of hours worth, sometimes four.
Here's the most recent 2-week period for grid-tied solar in Ontario:
View attachment 200061
Let's say we wanted to provide a constant 100MW during the day and overnight.
The first obvious problem is that production peaked at 333MW on Thursday but was only 38MW on Friday. Tuesday tapped out at 14MW from 478MW nameplate. This means that Tuesday's total production was only 68MWh or 2.8MW spread over the 24hr period, a 0.59% capacity factor
So, what kind of capacity do we need in this scenario to provide 100MW for 24hrs, which is 2,400MWh?
There are two ways we can approach this:
1. Use average capacity factor and assume that we'll be filling our storage with other sources when solar is unavailable. That could be gas and therefore counter-productive.
2. We use the worst case capacity factor and assume that the grid will just deal with over-production of what we can't store when production is higher.
Under scenario #2, we would need to use a period like Tuesday as our example, since we are trying to "firm" this 100MW, meaning it needs to be available all the time.
So, working that backwards, we'd need 16,950MW of solar (assuming no round-trip storage losses) to provide our constant 100MW; our 2,400MWh under Tuesday's conditions and peak output from the 16,950MW would be 496MW.
If we wanted to put some prices to these figures:
- Nanticoke solar (44MW), which is the most recent commercial project in Ontario, was $1.547 million/MW
- 16,950MW based on Nanticoke would then cost $26.2 billion
- This would occupy 150,240 acres; 235 square miles
- You'd need transmission capacity to handle full nameplate. Since that's roughly 3x that of the Bruce site, the largest power plant in North America, you are looking at several billion dollars for that as well.
Scaling Tuesday's data up to 16,950MW nameplate we get:
35, 213, 460, 390, 460, 496, 248, 106
So, a 65MW deficit at 9AM, but then no need for storage until 5PM. That's 1,565MWh required to cover when our solar isn't producing, which was captured from the surplus (above 100MW) during the day.
To cover this with batteries, that's $1.2 billion, pumped hydro should be less, but most pumped hydro projects (and batteries) are designed to have a much higher nameplate capacity because they aren't designed to provide 16hrs of storage. The smallest one proposed here in Ontario, Marmora, has a 400MW nameplate capacity and 2,000MWh, which could technically meet our needs here. We do not have pricing yet for this project, but if we assume its priced similarly to Meaford, that'll be $825 million.
So, our 100MW of firm solar would cost $27 billion approximately (CDN dollars) + transmission
An alternative spin on this: we could massively increase the storage duration and reduce the nameplate solar capacity, if we make the assumption that there's only going to be one "worst case" scenario day and that preceding and subsequent days would be better. 7 days of storage would be $4.5 billion. Not going to re-run the numbers, but based on Thursday and the subsequent Sunday/Monday/Tuesday, you'd need considerably less nameplate solar to make that viable.