To account for those rapid jumps in non-renewable demand, you need an energy source that can ramp up really fast on short notice, like hydrocarbon fire.
Even if reactors aren't as fast to react as a gas peaking plant is, perhaps batteries will soon be able to bridge that gap. The economics of batteries change greatly if you only need them to carry the load for tens of minutes for reactors to ramp up, vs needing to carry the load until the sun shines again.
The duck curve is a smoke screen and should only exist for a region that doesn't have solar to the west of it. Those solar farms in CA should be feeding users a timezone over to the east.
Any source can then feed into those stores, wind, solar, hydro.
If your power solution is involves hydrocarbons, it should be a closed cycle.
Or batteries and other storage methods. It looks like battery solutions could get cheap enough in the next 10-20 years to smooth things out and take care of storage for a few hours (see for instance http://news.mit.edu/2018/metal-mesh-membrane-rechargeable-ba... .. Donald Sadoway has some good talks on YouTube)
But yeah, for days that happen to have less wind/solar, I think the best thing is just keep gas power plants around. The CO2 impact for that will be minimal, the power plants are already built, and over time you could replace the fuel with synthetic gas made from renewables, which would basically be another energy storage mechanism. Cheaper renewable gas is something we need to make the world sustainable anyway.
The only unsolved problem is seasonal variations in colder climates. But those areas could import more trash and burn it (for both electricity and heat), like Sweden does. Norway has a ton of hydroelectric power, and is building more HVDC power lines to other areas of Northern Europe which will help with that area. Not sure what the solution for North America is though. But it's not nuclear. Having a nuclear power plant idle for half a year is the exact opposite of what you want. The plant is expensive and the fuel is cheap - if you build it you want it to run 24/7.
Only some unexpected breakthrough could make it feasible.
https://www.agl.com.au/about-agl/media-centre/asx-and-media-...
Does the hourly schedule variate day by day or can it be predicted?
If you can predict when and how much power you will need to produce then it can be ramped up slowly, I assume.
There are electrical dispatchers who are monitoring grid supply 24/7 and instruct plants on how much they are responsible for generating on a minute-by-minute basis.
Storage batteries, pumped storage, and CAES are examples of this, though simple raw thermal banking (hot water heating, typically) is an excellent way to suck up excess Joules or GWh.
1 GWh is roughly the energy required to heat a pool of water 1 hectare * 1 m by 86 degrees Celsius. This scales to multiple GWh by increasing area, depth, or both. Conversion to steam is also possible, though that requires more engineering (pressure is A Thing). Substrates such as molten salt have a lower heat capacity per unit mass, but can be heated to far greater temperatures.
Storage at the scale of entire US generating capacity for multiple weeks using molten salt thermal storage, even accounting for Carnot cycle efficiency losses (about 20-50% depending on specifics, 30% is a good ballpark) is actually a tractable-scale concept. Existing petroleum storage facilities are roughly comperable in size, though molten salt would require somewhat more robust facilities and insulation.
Whilst it doesn't have the net efficiencies of pumped hydro (exceeding 90% round-trip storage efficiency), pumped hydro lacks sufficient developable sites, and has significant environmental impacts.
The events you may have noted in news of "negative energy prices" are often failures of prediction -- unexpectedly high availability (more sun or wind), and unexpectedly low demand. Though "pay to take my power" sounds good, it's actually a sign of mismanaged resources.
There are occasional incidental factors -- sudden demand, or more often, equipment or transmission failures which require bringing additional capacity online, or shedding load to prevent under-voltage (and hence: over-amperage), or underfrequency. Grid power frequency is generally 60Hz in the US, 50Hz in the UK, and just for grins, both in Japan, on separate and noninterdependent grids, which made generation capacity loss following the Tohoku earthquake/tsunami and Fukushima incident all the more critical. Loss of synchronisation or deviation by more than a very small fraction from the nominal frequency is considered a Very Bad Thing. Viz the recent UK blackouts.
You can read more about it at http://nordicbalancingmodel.net.
I’d rather have way too much nuclear power than rely on any fossil fuels. If anything, energy demand seems posed to always increase.
What can't ramp quickly is the steam turbine connected to the reactor, but that's more of a design decision than a technical limitation. The German nuclear plants ("Konvoi" series) can in fact ramp faster than the German gas turbine plants, because that was a design requirement.
