I would prefer a mix.
Hydro is fine if there is enough water.
There have been periods of prolonged drought, some like the millenium drought in 2001 to 2009 lasting multiple seasons.
Tasmania, which has a large resource of hydro, in both 2007 and 2016 had severe drought that threatened to curtail power supplies.
Two things about this one really.
First is simply conservative operation within technical limits. I say that since the Tasmanian system has, using inflow data back to 1916, been 100% reliable if operated within its firm capacity rating. That it has on occasions failed in practice being ultimately a failure of humans rather than engineering.
4 incidents during the 1950's and 60's simply due to very rapid demand growth, consistently running at 10% per annum compounding, and construction struggling to keep up due to shortages of materials and labour. The system as it existed at the time was pushed hard, mostly coped but did fall flat on four separate occasions.
2007 simply because system load surpassed sustainable capacity in late 1997 and continued rising with no easy fix at the time for political reasons. By the time a fix was implemented, it was too late to recover storages before a serious drought turned up.
2016 simply because "just in time" prevailed over "just in case" as a management strategy, that being a fundamental aspect of how the market runs these days in all states.
Regarding the latter, well the companies get paid to generate, they're not paid to hold reserves "just in case", so that creates a conflict between the technical aspects of operation and the financial aspects of running a business. Bearing in mind that in Tas as well as Snowy, hydro might be government owned but it's run as a "private" business in practice and
required to make a profit. That's where it all went wrong, I know the story all too well. It was a failure of humans, it wasn't a failure of maths and science.
In the medium term though, what I'd do is very much a mix.
Nationally there's already a considerable fleet of fuel burning plant of various descriptions that could be used for deep firming. Long term that might not be acceptable on environmental grounds but it's there right now, it exists, and we're in no place to be getting rid of things that are perfectly useable.
So any hydro or hydrogen projects won't be the only things around anytime soon. They'll be part of the means of deep firming but by no means the only means and that does provide a very simple solution operationally.
Take the existing gas turbines etc and where possible add a second fuel, and don't use the same second fuel for all of them. So that is add diesel or add propane or whatever but the idea is we get an overall fleet with considerable fuel flexibility. That mitigates both physical supply and price risks.
From there it's straightforward to simply list them all on the basis of operational cost at any given time from cheapest to most expensive. That'll shift as relative fuel prices shift but it's not rocket science, it only needs a couple of people employed to keep on top of that.
So we have a list that, at any time, is cheapest through most costly to run and which as part of that includes any constraints - eg being able to run only 5 hours per day on gas, then having to switch to diesel, etc (and yes that's a real, actual constraint that exists at an existing facility).
For the hydro or other storage, it's then a case of not using actual cost, since that's effectively zero, but simply deciding where they need to be placed in that list. That is, if more water's available then place them toward the cheaper end, if less water's available then move them toward the more expensive end. Doing that scientifically based on forecast weather, current water levels, expected utilisation at any given point in the list (more about that in a moment) and so on.
Now here's the clever bit....
Since on most occasions when there's any need for it at all, only a portion of the deep firming fleet will actually be used. Only under the most extreme conditions will all of it run at once. That being so, that toward the cheaper end will do the vast majority of the work in practice and that makes the management of it reliable.
Eg real world example, Koombooloomba Dam in Qld, the Kareeya hydro scheme, is presently 87.2% full and that's despite heavy use in recent times. So it would sit at the very cheapest end of the list.
On the other hand Lake Echo in Tas is 44.1% full. Unless the BOM's forecasting a flood, it would thus be placed more toward the expensive end of the list. That's a human task, preparing forecasts and deciding where it needs to be put, and reviewing that on an ongoing basis - if the rain starts pouring down then move it accordingly on the list.
Noting that all wind and solar always goes in before any of the firming, it's always the first priority to be used.
From there, actually operating it is actually simpler than it sounds because we already have something known as the NEM Dispatch Engine (NEMDE). To be clear that's software running on a computer, not a physical engine of any sort, but that's extremely well proven to work - it's been in use for over 25 years now.
What NEMDE can and does do, is take into account all the constraints. Eg it takes into account transmission power limits and stability limits and so on. So it'll follow the preferred order
subject to compliance with all the other requirements. Eg it won't try to run the whole country from Queensland, it knows not to try that because the transmission ratings will be a constraint.
And yes it copes just fine with outages, there's already a team of people who simply put those in as needed and it works around that.
How that differs from the present is it's a shift from financial to physical. At present the basis is prices, which are simply numbers essentially made up by traders chasing profit, and that drives physical dispatch. That's how we end up with lakes drawn right down and so on, it's financial trading that's determining what's running at any given time and what's not.
Versus the alternative approach is basing it on real, actual costs and with a "team of stereotypical geeks" deciding where to put the hydro or other storage based on weather forecasts, known outages of plant and so on. A lot of number crunching but it's about the physical, it's not chasing profit. Pay said geeks for their skills and expertise, not based on how quickly they can drain a dam.
So that is, separate the financial from the physical. Pay (for example) Snowy to provide and maintain the assets and use those assets as required for the overall operation of the system. Don't pay them only when they let water out and incentivise them to do so unnecessarily. Remove that incentive to run storage down and replace it with a strictly science-based approach.
As a practical example of this working, Origin Energy owns the Quarantine generating site in Adelaide. All up there's 5 gas turbines at the site, all using the same gas, sending electricity out on the same lines and owned by the same company. So whilst Origin will certainly compete financially against other companies, and that may lead to irrational outcomes, they're not going to compete within those 5 turbines at that one site.
That's relevant because they're not all the same. 2 x 29 MW are the most fuel efficient, the single 128 MW unit is the second most efficient, and the 2 x 24 MW are the least efficient. Which gives rise to the following output over the past 12 months:
The 2 x 29MW units comprise 24.8% of capacity at the site but generated 74.8% of total output.
The 128MW unit is 54.7% of capacity and generated 23.5% of total output.
The 2 x 24MW are 20.5% of capacity and generated 1.7% of output.
That's a classic example of what happens when it's done that way. The lowest cost plant gets run hardest, the highest cost plant is used only at the extremes. Bearing in mind this is a peaking plant so it's a good example, it's a plant that would be retained long term for deep firming purposes. Capacity factor on the 29MW units was 18.0%, for the 128MW it was 2.6%, for the 24MW machines it was just 0.5% over the past 12 months.
Apply that same logic across the whole fleet and, so long as it's based on real actual costs and a proper, scientific approach to deciding where to put the hydro in, then I've zero concerns about it working. I'd be willing to place a very large bet on that, as an approach it's a well proven concept.
The key is basing it on real engineering criteria, separating that from someone trying to make money out of it. That doesn't mean nationalisation, it just means changing how the market works. Get paid to provide the facilities, rather than being paid only when they run which incentivises draining them out.