Battery storage: better at peaking than gas in South Australia

Print Friendly, PDF & Email

Existing  gas generation in South Australia is expensive, inefficient, highly polluting, and is controlled by just two market players. Battery storage is emerging as a much smarter and cheaper option

An alternative to building more gas peaking plants in South Australia – which would take years and require hard-to-get-gas – is to build a lithium battery storage plant to act as a peaker.

We don’t have access to formal quotes as to the price but our industry contracts suggest that around US$900 KW is achievable installed. On that basis we think a lithium peaker might have earned a best case return on equity of 16% last year. (See Figs 5, 6 and 8 below).

There are no barriers to entry for battery speakers, so in the long run returns will go to the cost of capital. However, over the next two-three years after Hazelwood closes, prices in South Australia could well go higher, even assuming Pelican Point gets back to running flat out.

A battery would also help the gas generators. It would charge when wind supply is high and pool prices low. This would reduce the losses gas generators incur in those conditions.

Who knew South Australia was so interesting

To recap. Last week we went through some facts that essentially showed that despite the storms and the heat at least two  of the blackouts in South Australia, including the Statewide blackout need not have occurred if AEMO had done its job better.

We estimate $350 million of damage was caused by the storm related blackout last year and it wouldn’t have happened if the wind farms had had the same “ride through” settings as installed in Europe a decade ago.

We are not going to talk about politics today, but contrast the publicity AEMO’s latest views on the matter have received compared with the ideological fight of “price/security v the environment” at the time of the blackout.

Instead, we want to focus on the challenge posed by wind’s share of South Australian generation. Over the past 12 months (which does include the last days of Northern Power station) wind has supplied 38% of South Australian demand.

It’s over 42 % in Denmark and that’s not a problem, despite limited imports and exports in that country, because there are enough other power plants that can flex up and down.

In South Australia the problem is two fold.

  1. When demand is really high the State is actually short on generation capacity.
  2. When wind generation is high and demand is low there is too much wind.

The two problems are related because when there is too much wind gas generation is unprofitable and price can be negative.

The following chart over the past 12 months shows demand in South Australia after subtracting wind generation. Its arranged in percentiles. The horizontal axis shows the percent of demand less than a given “y”.

Figure 1 Demand in South Australia ex wind. Source NEM Review

Figure 1 Demand in South Australia ex wind. Source NEM Review

Gas generation wants to run at a constant output

Demand is negative around 5% of the time and < 100 MW around 9% of the time. Obviously it’s not much fun being a gas generator when net demand is very low. Your choices are to export to Victoria where you will compete with a much lower variable cost brown coal generator or to close down.

Looking at the other end of the curve there is a 600 MW increase in the last 1%. We’ve made a separate graph of the last 1% and it shows that most of that 600 MW occurs in the last 0.4% of time which is about 70 half hours for the year.

Figure 2 Highest 1% of non wind demand. Source: NEM Review

Figure 2 Highest 1% of non wind demand. Source: NEM Review

This is a traditional problem for electricity supply. It’s expensive to put 400 MW of capacity in place to satisfy 35 hours of demand.

Traditionally that’s been done with jet fuel/diesel  type generation because the gas supply can’t always be guaranteed at short notice. We only focus here on the generation issue, but it’s worth noting that in South Australia distribution and transmission of electricity is expensive and that system too has to be sized for those top 35 hours.

Gas generation in South Australia is very inefficient

Most of the gas generation in South Australia these days is from he Torrens Island TIPS B and Quarantine power station.

The heat rate of these is about 111-12 GJ  MWh, This is very wasteful of gas and carbon inefficient.

Figure 3 Heat rates of manin Sth Australian gas generators. Source: Frontier economics

Figure 3 Heat rates of manin Sth Australian gas generators. Source: Frontier economics

Gas generation doesn’t really want be ramping up and down all  the time. Gas turbine efficiency is less at “half power”.

Gas generation in South Australia is an oligopoly

Not only is the gas generation inefficient with carbon emissions on average close to coal generation but most of its market share is split between AGL and ORG. ITK is not suggesting that either company takes advantage of this but we are suggesting that consumers generally do better when there is more competition.

Figure 4 Gas generation shares last 12 months . Source: NEM Review

Figure 4 Gas generation shares last 12 months. Source: NEM Review

Transmission, gas or batteries?

Is gas the answer?

So why build a 400MW gas peaker  capital costing say $500 m (if your lucky), plus fixed operating costs including rentjng gas transmission capacity for 70 hours a year?

Why force a combined cycle gas generator into supplying Victoria when wind generation exceeds South Australian demand (5% of the year or 400 hours a year)

Transmission – would benefit if the proper planning studies were done

More transmission may well be part of the answer and is ultimately needed to take advantage of the portfolio effect of wind (wind output in NSW or QLD is not so correlated with wind in South Australia) so that the variability of total wind generation becomes less.

However, transmission itself is expensive and takes time to build. To demonstrate the value of transmission basically requires showing a net benefit to consumers (those of the source as well as the destination). Wind from South Australia would flow to NSW at times of excess supply there and vice versa. Studies to demonstrate that value don’t exist to our knowledge or certainly haven’t been publicly released.

However, transmission studies don’t generally look at renewable energy portfolio impacts because like everything else the studies are done without the benefit of NEM wide forward model.

Lithium storage

Is lithium storage economic? One answer is that if it was someone would have done it already. So lets change the question round and ask why hasn’t it been done?

  • skepticism from say AEMO.
  • industry views towards lithium storage change only slowly
  • you could argue the incumbents aren’t that incentivized.
    • Networks aren’t allowed to access market revenues or can’t combine network and market revenues into one business unit
    • Existing generators of the gas variety are doing well
    • Wind farm owners particularly ones such as Infigen which have merchant (uncontracted) outuput would seem the most obvious case.

Lithium battery as a peaker

Lithium batteries can provide the same service as a peaking gas plant only better.

  • In addition to stopping high prices they can also help to prevent negative prices by absorbing power when wind generation is high relative to demand.
  • Battery capital cost is comparable to peaking gas capital cost. The last gas peaker built in the NEM was the Mortlake station in Victoria and cost $660 m for 550 MW. It’s a bad example as it ran into union issues in Victoria that made it expensive but actually took 4 years to build. We think Lithium storage is right around the $1.2 m/MW mark.
  • Operating cost is much lower
  • Can be built in less than a ¼ of the time.
  • Can be built in modular 1-50 MW units allowing maximum flexibility for deployment. Scale advantage is achieved when the battery can justify its own building centralizing airconditioning and fire suppression and avoiding the need for containerization.

We think lithium storage can be built and installed for about A$1.2 m/MW. That’s for a notional 4 hour system. In this world though it probably doesn’t run for 4 hours. Its probably half or ¼ full most of the time (it has to be available for charging as well as discharging)

We think it should aim to earn a return on equity of about 8%. Here’s some capital and cost of capital assumptions.

Figure 5 Battery assumptions. Source: ITKe

Figure 5 Battery assumptions. Source: ITKe

A return on equity of 8% on an equity investment of $6 m is about $0.5 m after tax. We work backwards from that to get to the required EBITDA (earnings before interest, tax and depreciation) of $1.6 m.

Figure 6 Peaking battery P&L. Source: ITKe

Figure 6 Peaking battery P&L. Source: ITKe

Could we have earned $1.6 m of ebitda last year. Sure Can (whoops can’t make any Rolf Harris jokes)

The Figure below shows the distribution of pool prices in South Australia in the 12 months ended Feb 17.

Figure 7 South Australian pool prices. Source: NEM Review

Figure 7 South Australian pool prices. Source: NEM Review

Using that data we formulate some decision rules. These include discharge battery when price greater than X and recharge when price less than Y. Using seemingly conservative rules designed to not run for more than 500-600 hours per year we see our battery peaker earning around $3 m over the past 12 months.

Figure 8 Peaker notional gross profit Source: NEM Review, ITKe

Figure 8 Peaker notional gross profit Source: NEM Review, ITKe

These numbers look to be about twice what’s required but there are some caveats.

  • We assume that the battery has enough energy capacity to take advantage of all those hours. If all the high price events occurred at the one time the battery would run out quickly. Same goes for charging.
  • That is just for last year and assumes the battery that is installed has no impact on the price. The reality is that if you install the battery with an expected 20 year life and the economics are as good as suggested in Fig 8 then there will be lots of batteries installed and the return will go to the marginal return. But hey that’s why markets do work.

David Leitch is principal of ITK. He was formerly a Utility Analyst for leading investment banks over the past 30 years. The views expressed are his own. Please note our new section, Energy Markets, which will include analysis from Leitch on the energy markets and broader energy issues. And also note our live generation widget, and the APVI solar contribution.  

  • George Michaelson

    Since this is well within the budget of a number of interested groups whose primary goal is low carbon intensity and not profit, It feels like a ‘money where your mouth is’ moment. I don’t think kickstarter is the vehicle, but surely there is some funding agency which could get over the line on this?

    • Andy Saunders

      ARENA… plus CEFC

  • alexander austin

    David. Great article – I published on a similar topic last week – see

  • Malcolm M

    You have overlooked the requirement by AEMO of having at least 3 large (60-100 MW) generators operating at all times in the SA market for system stability. This means that even when SA is exporting wind power, it must still be generating ~180 MW of gas power. The interconnector capacity is only ~800 MW. If the ~180 MW of gas generation were instead changed to 180 MW of battery, the capacity to export wind energy would increase by ~180 MW.

    Furthermore, some of the inter-connector capacity is reserved for frequency control and stability (FCAS), say 100 MW. This is because the gas generators can’t respond quickly enough to minor fluctuations in demand and renewable supply. But with the much quicker response of batteries, all of the FCAS could be supplied within the SA market, allowing the inter-connector to be deployed at is fully rating.

    The most logical investors in such a battery system would be those with solar farms, to compensate for the precipitous drop in production that occurs when a cloud passes over. There could be multiple revenue streams, such as FCAS, and buying and selling based on market forecasts.

    • JeffJL

      I have thought for some time that storage for current FF generators would be a no brainer. Instead of having to ramp your production up and down you could run your generator at its most efficient capacity dipping into storage when the market needs more and charging when the market needs less. Must be more money in running the generators inefficiently.

  • suthnsun

    FCAS revenues will be additional?

  • alexander austin

    One question I have is how transmission charges interact with battery economics? Under the current rules wouldnt a grid scale battery need to pay transmission costs on the power it uses to charge (unless it was directly connected to a generator)?

    • David Mitchell

      That would be my assumption too. I ran a ruler over the SA wholesale numbers for 2016 as well and got a similar result, but if you have to pay for electricity delivery, the return is not so attractive.

    • Chris Schneider

      The smart money would place these near variable loads though (wind and solar) and then transmission costs would be negated.

  • JeffJL

    Another reason for there being no battery peaker plants would be the rapid falls in price of the batteries. Investment decisions of this magnitude have lengthy assessment processes. If you were putting in a battery peaker now you would have started the assessment process say five years ago when batteries were three times the price and not nearly economical.

  • Mark Diesendorf

    David, agreed, batteries have a big future as ‘peakers’, but you have overstated your arguments against open cycle gas turbines (OCGTs), which should not be dismissed until batteries are less expensive and more widely accepted. Contrary to your statement, OCGTs can be ramped up and down easily and quickly, as occurs when you fly in a jet aircraft. OCGTs cost US$800-900/kW and can burn almost any liquid or gaseous fuel, including biofuels, hydrogen and ammonia from renewable sources. When used as peakers in a predominantly renewable electricity system, OCGTs only have to be operated for a few percent of the time, filling occasional gaps in supply when demand is high and simultaneously wind and solar PV supply happens to be low. So, their annual fuel cost is low. Like future lower-cost batteries, OCGTs can play the role of reliability insurance with low premiums.

    • Andy Saunders

      Mark, yes, there should be a distinction between CCGTs and OCGTs as you write.

      I’d also say the article comment “Gas turbine efficiency is less at “half power”” is somewhat wrong. Modern turbines have greater turn-down efficiency than previously, and in any case, two turbines won’t be dispatched at half capacity – only one will at full capacity. His point is somewhat correct, efficiency always decreased as the power output is throttled back, it’s the degree of efficiency loss which is somewhat exaggerated.

      • Ian

        No it isn’t. The ramp rate and the fuel compatability are. An OCGT at running temperature will take anywhere between 3 and 8 seconds to go from 30% to full load if it is a two shaft machine. A single shaft machine takes up to 0.5 seconds, slightly more if the IGVs have to open. CCGT can take minutes due to the process lag. Turbines are only able to take certain percentages of hydrogen, ammonia and CO2 too. The flame speed of hydrogen is too fast and ammonia is too slow. Ramgen (aka Siemens nee Dresser-Rand) are the only known brand with the combustor design capable of taking pure hydrogen, in their prototype machine.

    • Peter F

      Mark, I agree with most of your benefits of OC gas turbines, part load efficiency is not one of them. Even the best Siemens and GE turbines are only 25% efficient at 40% power and are very unstable below 40% power. That is at ISO standard conditions (15C) not 42C where efficiency will be around 20%.
      Also at such loads they must be ramped relatively slowly., from off to 30% power takes 7-8 minutes. Batteries can go from full charge rate to full discharge in less than a second.
      The benefit of a battery/gas turbine combination is for periods of more than an hour of low wind low sun where the battery might run out of puff or peak half hours where both work together.
      I don’t know how to calculate the benefits but in order of preference I would place batteries
      1. at the load, to minimise T&D investment and losses
      2. in the distribution grid for the same reason but to aggregate excess local solar and gain economies of scale in plant and installation costs
      3. at the generators to maximise generator efficiency

  • Andy Saunders

    David, would be interesting to see your views on flow batteries instead of lithium…

  • Brad Sherman

    Mark and David,

    Do you see any prospect for CSP to provide dispatchible power and frequency services? I’m thinking of Solar Reserve’s Crescent Dunes plant in NV and their proposal for 2GW ( assume the technology scales) with 24h full capacity storage that I read somewhere could sell into the grid at 8 cents/kWh. I’d be interested to learn how much subsidy may be embedded in the 8 cents/kWh. Their technology seems to be reasonably succesful based on something I read in Oct after 1 year’s operation where the plant exceeded both design efficiency and contracted power sales.

    It just seems to me that these things might be directly swappable for the old coal-fired stations, especially in areas with good solar resource. That would take a lot of the steam (pun intended) out of the government’s arguments, wouldn’t it.

    • David leitch

      CSP has a lot of potential but also some significant disadvantage.
      1. Takes years to build.
      2. Has to be built in relatively large lumps of capacity
      3. Mechanical operation/coordination/focussing of the mirrors has not generally been all that good so that my understanding is that existing CSP plants often dont achieve their designed capacity utilization. If you google around on Ivanpah, perhaps a bad example, you will get the idea.
      Like everything else if enough of these plants were built they would do much better.

      • Brad Sherman

        I see that Crescent Dunes took 4 years from breaking ground to delivering power, i.e. completing all commissioning, etc. It’s selling its power at 13.5 cents/kWh and has a federal govt loan guarantee. I assume a loan guarantee is not a subsidy unless the firm in question goes broke, is that right?

        Is 4 years a long time to build a power plant? Is there a typical time to build, say, a 100 MW OCGT plant?

        I don’t think it’s fair to tar CSP with the Ivanpah brush. It uses water, the site caught on fire last year, and there have been other site-specific rather than technology specific problems. A lot of learning has happened since Ivanpah – which I think was truly in the vanguard of CSP when it was constructed. I’ve been persuaded that the molten salt technology such as that used by Solar Reserve is preferable – it doesn’t seem to require busing gas to get the system going in the mornings. I’m guessing Solar Reserve also learned from other CSP experiences regarding how to defocus the mirrors to minimise harm to wildlife (no focal point with > 4 times full sun intensity).

        How much does a recently deployed grid-connected battery-based system have to charge per kWh to achieve a payback period of its design life (I assume about 10 year)? In your example above, is it ~ 8 cents/kWh neglecting interest charges and assuming no O&M expenses?

        I gather China has something like 40 CSP plants under construction or advanced planning, mostly in the 10-40 MW range. The Solar Reserve system seems to be made of ~110MW units.What do you consider a large lump of capacity?

        How big a generating capacity is required to black-start the grid in SA, for example? How much power and what duration of supply are required to allow the other generators to synchronise? I recall being told the Fitzroy Falls hydro PS (~200MW I think) could black start the NSW grid.

        I’m think about CSP more in terms of being a replacement for fossil-derived baseload power at night rather than as a peaker. Presumably, the large chunks of capacity are less of an issue in this regard. Certainly battery technology seems the logical choice to deploy alongside wind and PV systems.

        Last question (and hopefully provoking another article by you!) – The design of the Australian Energy Market seems to be adversarial in nature. Companies bid against each other rather than collaborate to provide a more optimal emission per energy objective. In fact, I’ve been told by some in the generation industry (while standing in their control room) that they do on occasion deliberately structure their generation to harm competitors. It seems to me that everyone will want to sell into the market during periods of peak demand. These periods coincide with peak PV and wind generation. Storage, like pumped hydro, appears to me to best suit surplus large baseload capacity and is likely to be used to justify the roll out of new fossil plants rather than buffering wind/PV – certainly when I worked for the hydro division at PG&E a long time ago we used PSH to meet peak demand rather than to absorb renewable energy generated during periods of peak demand (although RE was very much in its infancy back then). How do the rules governing the NEM need to change to facilitate more rapid introduction of energy storage?

      • Brad Sherman

        Because I was asking about replacements for existing coal-fired plants I think generation in big chunks of capacity were implicit in my assumptions. Big chunks of capacity presumably make sense for particular types of demand and the fact that the transmission infrastructure is already in place must surely help the economics.

        I think Ivanpah is a bad example and the design differs considerably from Crescent Dunes. I consider Ivanpah a stepping stone (the first, perhaps) in the evolution of CSP. Apart from the leak in the storage tank at Crescent Dunes, it seemed to be performing well – although I haven’t heard anything about its performance since the leaking storage tank was addressed.

        Must be a challenge having to repair a leaky weld in a tank with ~ 500 deg C on the other side. I wonder if getting the temperature down to something more feasible to work with was responsible for the longer repair time. Again, it’s the first go at it so things are bound to be slower as we learn what works and what doesn’t.

  • Peter F

    I think you are on the right track, however
    1. Your battery cost estimates are probably too high. In the UK they got 200MW of batteries for A$150m or A$750k/MW.
    2. In Ireland they say a 10 MW battery is providing equivalent FCAS services to a 100 MW gas turbine. That reduced costs of spinning reserves. I think I read that its typical state of charge is around 60% because the cost of wasting excess wind is far less than the cost of short term firing of a gas turbine.
    3. If a substantial share of the batteries were distributed in substations they could reduce peak transmission losses and more efficiently capture excess rooftop solar so reducing overall system energy use.
    4. If they were large enough they could make a significant contribution to grid strengthening particularly on remote parts of the grid.
    5. AS JeffL says below they can actually help the efficiency of CC gas and even OC gas by allowing to run near full load for a greater share of their operating time.
    6. The economics/stability/efficiency of gas turbines running at low loads is quite complex and their response to step or pulse loads is very uncertain (see Quarantine during the blackout) close coupled batteries ameliorate these concerns again making more money for the operator reducing gas consumption while toughening the grid

    All these services have significant value to the grid and save money for everyone. The question is who pays in addition to the arbitrage model above. A vertically integrated grid in SA faced with current gas prices would have no trouble at all capturing all these benefits and would probably show 10-15% ROI

  • tonyk

    David’s battery price wrong (too low) by a factor of 2 or 3.
    David says
    ‘We think lithium storage can be built and installed for about A$1.2 m/MW. That’s for a notional 4 hour system.’
    This is $300/kwhr.
    Large high voltage storage systems closer to $800/kwhr when you include inverters, fire suppression, buildings etc etc.

    • David leitch

      tony. You may be right, however the numbers I quoted certainly included the inverters. The cost per unit will be lower using a building than using containers provided the scale justifies it. In the Utility side of the battery business my understanding is that the industry talks power i.e. MW not energy MWh. Costs are quoted per MW with the hours a secondary discussion. We are all learning about this topic and certainly I hope to improve. My number came from what I believe to be a very informed source, albeit with no detail.

      • tonyk

        My sources say Tesla batteries >$A800 / kwhr installed.
        Have been supplied for large installations (80MWhr in California)
        currently quoting $1.2 mill for 2100 kwhr
        plus gst
        plus installation
        minus derating (this is the brochure rating need to look very closely at how this defined ).
        plus any costs to get up to transmission voltages (Tesla units only have 415 volt inverters.
        minus any volume discount.
        None-the-less there will be places in grid where these costs viable
        where you avoid grid upgrades.
        Batteries not cheap but a few hundred km of high voltage line not free either.

  • Ian

    Pelican Point has a heat rate of 7.5 (LHV) at ISO conditions if in CCGT mode. The problem is that it is rarely operating in that mode when having to deal with peaking. In OCGT mode it will have a much worse heat rate, much like TIPS or even worse.

  • “$1.2 n/MW … for a notional 4 hour system” is not equivalent to a peaker gas turbine or any other fossil fuel backup when your intermittent renewables are out for days at a time.

  • Ren Stimpy

    Is lithium storage economic? Yes, by many accounts.

    If the government had a SET or storage energy target, whereby the cost to the consumer was subsidised by say 10%, then it would certainly be economic. The subsidy should be conditional on participation in grid reliability programs such as grid load shedding (remotely switching homes and businesses to their own battery power to reduce load on the grid). The taxpayer would get a huge return in the form of a more reliable yet less gold-plated grid, as well as a lower requirement for baseload generators.

  • Great piece David. I agree that battery is more more likely than gas as the next peaking plant.

    If the SA government wants to do something productive, they should commission a 100MW battery facility. Let the taxpayer be the owner.

    This project should come with a couple of restrictions.

    It should not be owned by or allowed to contract bilaterally with either of AGL or ORG. Adding another player to the poker table would be good to erase any perception that these two are manipulating the market (not that I would ever suggest such a thing…).

    Secondly, it should be run to optimise profitability. If it’s a good investment, that will create a funding pool for expansion and preserve price signals for other market participants.

    It’s quick, relatively cheap and would have a significant impact on the stability of the SA grid.


    Dave P.