Has Queensland overpumped its pumped hydro plans?

The extremely tentative conclusions from this note are that pumped hydro and associated transmission are a large part of the Queensland plan capital cost, much bigger than Snowy 2.0. Secondly, it might be that Queensland should be building even more wind than hydro – and then it might need a bit less firming supply.

Using AEMO estimates, which are much the same as what Queensland proposes to do by 2035, I would conclude there will be times when much of the proposed pumped hydro capacity is required, but that most of the time much of it will sit idle.

On a power basis, it’s likely that a distributed battery system would have a similar cost and, in my opinion, would be more resilient with fewer social license issues. On a seasonal energy basis, the pumped hydro with 24 hours storage is close to an order of magnitude cheaper, even after allowing for the transmission.

However, much less storage duration is required in the AEMO Integrated System Plan because the highest weekly demand for storage in Queensland is not in winter but, at least in this simulation, in late March. Hydro and southern state production should still be quite good then, and so imports can also help.

The Queensland plan is a very big plan and this note is just an attempt to unveil a thing or two. 

Breaking up the cost

Using extremely broad and sure-to-be-inaccurate rules of thumb and industry knowledge, our capital cost analysis of the Queensland Super Grid is:

Figure 1. Estimated cost of QLD super grid. Source: ITK estimates

These costs are shown on an “overnight” basis. That is, all the capitalised interest accumulated during construction is ignored. Similarly, development costs such as land acquisition and EIS and bio diversity for the pumped hydro is equally ignored.

Most of the transmission cost is shown within the pumped hydro segment, as that is how its discussed within the Queensland SuperGrid Blueprint. The transmission costs are as disclosed within that document. All other numbers are crude estimates.

The biggest uncertainty is the pumped hydro, and so that analysis follows.

Pumped hydro v distributed batteries

One of the differences between the AEMO ISP and the Queensland SuperGrid plan is that the former emphasises household storage, whereas the latter has 7000MW of 24 hour pumped hydro.

In essence, one plan has a consumer-driven distributed solution and the other has a state planned and owned infrastructure with “old style” concrete and transmission.

There are many ways to think about these different approaches – and nor are they mutually exclusive. In my view, a clever government would do as much as it can to encourage household and neighbourhood batteries.

The Queensland government does own the distribution networks and could absolutely lead a revolution in electricity distribution technology, if it was so minded.

The jobs are different. Distributed batteries provide local jobs in cities and there is a strong emphasis on software. Pumped hydro projects involve lots of carbon intensive cement, rocks, building big dams and the associated transmission infrastructure. The jobs in the Pioneer region are well regional.

To start with, let’s go back to the map provided by the DPIE.

Figure 2. QLD super grid. Source: QLD Govt

It’s basically 1000km from the proposed Pioneer site to Brisbane. By contrast, the much commented on Snowy to Sydney distance is about 500km.

The Borumba site is just 175km from Brisbane and a cosy 65km to Noosa.

On our numbers the total cost of the pumped hydro and associated transmission is about $19 billion and it’s a 13 year program to get it all done. I work pumped hydro costs on $1.93 m per MW and $0.103 m per MWh of storage. There should be savings on bigger storage because the volume of storage is very levered to a change in the area of the storage walls. 

Figure 3. QLD pumped hydro cost estimate. Source: ITK back of the envelope

Because the incremental storage is relative cheap (I used $0.103 m/MWh) the total cost including transmission works out relatively low, in terms of MWh, but then it will need to be because the revenue opportunity for a 24 hour plant is far smaller than for say a 4 hour battery that can expect to run 3-4 hours every day and also provide system services. By definition, a 24 hour plant won’t run for 24 hours a day and in fact can’t run more than 10 hours a day consecutively.

Out of interest, what would 7000 MW of Powerwalls cost? To start with what would 1 Powerwall cost? I was astonished, in all honesty, to read recently over on Solarquotes that a Powerall can cost $A19k, post installation and including the basically essential gateway. Pass the BEX and lead me to a good lie down. Makes me glad that after waiting in vain for four years for the price to come down I bit the bullet and ended up at about $13k installed with gateway.

Figure 4. Powerwall cost history. Source: Solarquotes.com.au

The higher prices are, of course, at odds with the basic expectation of learning rates driving down prices and costs. In the short term, this has been turned on its head both by brand, Tesla has a large brand price premium it can charge because demand exceeds near term supply, and because underling costs, primarily raw materials costs have increased.

The best estimate of lithium content I’ve seen is that there are 850 g of lithium carbonate per kWh. At current prices that’s about $1250 of a Powerwall cost at US74k/t  and of itself could not possibly explain the price increase. Of course, there are other material price increases, not just lithium – but lithium is the largest cost driver from the materials piece.

A Powerwall provides 13kWh of storage and has a max sustainable discharge of 5kW or, expressed in utility terms, 5kW/2.75 hours.

Using what I regard as a better long-term figure of $13,000 for an installed Powerwall with gateway box then it would take 1.4 million Powerwalls to provide 7GW of power and that would cost about $18 billion – broadly comparable to the pumped hydro cost. There would also be many job opportunities but they would be jobs in the cities, installing the Powerwalls and writing the software to manage them.

In any event, to provide the 168GWh of energy requires 13 million Powerwalls and costs $170 billion for something we don’t need to spend any time on. And this begs the same question that I grapple with more and more: What is the value of longer duration storage? How much is needed?

According to the ISP, not much really. As a reminder, this was their modelled output with the Step Change scenario inputs.

Figure 5. ISP least cost QLD storage total and type. Source: AEMO

Out to 2045 the ISP only wants to build 1GW of long term storage in Queensland.

In terms of production, and this is what counts for revenue and profit  the numbers are small. 

In the year 2035

I’ve chosen to focus on 2035 because that is where the Queensland Super Grid numbers go to. However, in 2035 the cumulative build in Queensland exceeds that modelled by AEMO for the ISP step change and it takes until about 2039 for the ISP model to get to the Super Grid renewable capacity.

Figure 7. Installed utility solar and wind. Source: QLD Govt, AEMO

Another way of looking at this is to look at a firming duration curve. However, this is not straightforward, as there are both duration and size components.

In general, we measure firming demand as demand – VRE supply where VRE = the sum of wind, solar and rooftop.

A second point, one often overlooked when thinking about demand, is that pumped hydro and batteries also serve as the demand for the excess supply of wind and solar that typically occurs in the middle of the day or when its winder than usual. Pumped hydro and batteries provide value to that excess supply.

In terms of the average day the Queensland 2039 shape looks as follows:On average, 3.3GW of firming supply are required (max 5GW) between 6pm and 7am totalling 47GWh. This is before any allowance for existing hydro, interstate imports/exports, or anything else.

Equally to a first approximation the total VRE supply = demand so there is enough excess supply in the middle of the day to manage the overnight load.

But, but, but, that’s just the average and if there is one thing we know about averages it’s the joke about the three statisticians shooting rabbits, one missed to the left, the next to the right and third one then shouted “got him”.

Using a 48 half hour *7 rolling total (ie weekly total) shows that in FY2039 there is not quite enough renewable energy in the AEMO model, at least as we calculate the output, and it’s broadly equal to what AEMO shows as the totals. It’s not enough to allow the surplus VRE that is the lowest cost way of reducing the firming demand.

So looking at the daily and weekly firming energy totals shows that in one or two weeks as much as 600GWh of firming will be required, in total, and that is net of the VRE generated in those weeks. Again, this chart is from 2039. 

Figure 10. QLD daily and weekly firming demand FY39. Source: ITK

My Pandas/Matplotlib skills are still a work in progress. So the daily chart shows that in no single day does the demand for firming get as high as the pumped hydro capacity but that the worst weeks will still require some gas and the existing hydro. 

The worst weeks, though, don’t happen in winter, they happen around March, at least in this simulated year and at a time when imports from other states can help. 

To me the data illustrates the value of a well connected NEM.