The South Australia region of the National Electricity Market has become a focal point for those taking an interest in electricity industry transition, as it has a much higher share of variable renewable generation than any other region in Australia.
Contrary to the ideas promulgated by some opponents of wind generation, most of the substantial investment in wind farms in South Australia is not being paid for by taxpayers or electricity consumers in South Australia, but was and is being built under the nationwide Large Renewable Energy Target.
Under this federal policy, the cost of renewable energy certificates is recovered through retail sales of electricity. Late last year the Australian Energy Market Commission estimated that during 2016-17 the LRET added 0.8 cents/kWh to electricity tariffs, which is less than 4 per cent of the tariff.
This is offset by the LRET putting downwards pressure on wholesale prices, as found by numerous analysts. Even discounting this effect, 4% is a reasonable price to pay for what is currently the most important national policy for energy industry development and reducing Australia’s greenhouse gas emissions.
The largest wind farm in South Australia, Hornsdale (also the location of the Tesla big battery), is being built under the ACT’s feed-in-tariff scheme, not under the LRET.
The owners of Hornsdale were able to contract with the ACT Government at a price below $80 per MWh because of the financial security provided by a 20 year contract with the ACT Government.
Between January and August this year the owners of Hornsdale were actually paying the people of the ACT, because the level of wholesale prices in the NEM was often higher than the contract for difference price agreed with the ACT Government.
There are two main reasons why South Australians have had a wind farm boom in their state. First, there is a lot of wind close to the NEM transmission grid, which means that construction costs are low, and revenue is high. Second, South Australia’s government has been welcoming and supportive of the wind industry.
Over the same period, governments in Victoria and Queensland, particularly, were hostile and obstructive to wind farm investors. That has now changed in both states.
Wind accounted for 52% of large-scale generation in South Australia in September and 45 per cent in October. If rooftop solar is included, the total renewable share of generation was 55 per cent and 49 per cent, respectively.
This can be seen in Figure 10 (above), which plots average daily output by wind and gas generators, as well as output from rooftop solar.
Figure 11 (below) shows the same numbers but in stacked format, which makes it easier to see that trade in electricity through the two interconnectors with Victoria shifted from net imports up to June this year to net exports in every month since July.
This means that wind generation was larger when expressed as a proportion of electricity supplied from the grid to consumers in South Australia – 57% in September and 50% in October.
Figures 10 and 11 also show that the gas share of South Australian generation during September was just a few percentage points less than the wind share.
If generation by rooftop solar is also included, total renewables, i.e. wind plus solar, accounted for a record 55 per cent of all electricity generated in the state during September.
After adjusting for net exports of electricity to Victoria during the month, renewable generation was equal to 63 per cent of electricity consumption.
For rooftop solar, although South Australia does not lead other states in installed capacity as it does for wind, it is the leading state in terms of capacity relative to average demand for electricity. Rooftop solar therefore makes a larger contribution to total electricity consumption than it does in other states.
Over time, this is having an effect on the operation of the electricity supply system in the state, mainly through its effect on patterns of demand, that is no less profound than the effect of wind generation.
South Australia is experiencing profound changes in its patterns of demand. These are summarised in Figure 12.
Total consumption decreased from 2010-11 onward, as did annual peak demand. Annual minimum overnight demand has stayed roughly constant, but annual minimum demand in the middle of the day has fallen steadily, and since 2012-13 has been lower than the overnight minimum.
This significant change is largely attributable to rooftop solar generation. In 2007-08, nearly all of the top 2% of 30 minute demand periods occurred before 5 pm local time, many of them well before, when the sun was relatively high in the sky.
In 2016-17, most of the top 5% occurred after 5.30 pm local time, some well after, when the sun was much lower. Rooftop solar generation has had the effect of both lowering peak demand on the grid and pushing it to later in the day.
No wonder that South Australia is more advanced with trials of demand response than other states.
Changes in demand are also shown by plotting demand in each hour of the year in decreasing order, called an annual load duration curve.
Figure 12 shows load duration curves for electricity supplied from the grid, i.e. excluding rooftop solar and other embedded generation, in South Australia in 2007-08 and in the year from November 2016 to October 2017 (hence forward termed 2016-17).
It also shows the curve for 2016-17 if generation by rooftop solar is added to generation supplied through the grid. Figure 13 focuses on the top 10% of demand intervals.
A number of interesting points arise from these figures.
First, rooftop solar generation, which was virtually zero in 2007-08, has been responsible for a large part of the apparent fall in consumption between 2007-08 and 2016-17.
The remaining part of the fall reflects (the gap between the red and yellow lines) reflects an actual reduction in total consumption of electricity. In 2007-08, total consumption of grid electricity was 14.-0 TWh, while in 2016-17 it was 12.5 TWh, but 13.5 TWh when rooftop solar is added..
Second, while average demand has fallen, the annual 30 minute peak grid demand was almost the same in the two years. In 2007-08 the median 30 minute demand was 1,615 MW, equal to 52% of annual peak demand. In 2016-7 the corresponding figure was 1,391 MW, equal to 45% of peak demand.
Third, rooftop solar has already substantially reduced grid demand. 2016-17 peak demand including rooftop solar was 7% higher than the 2007-08 peak. In other words, rooftop solar has made a very significant contribution to reducing annual peak demand for electricity.
However, because the peak has now been shifted to just before sunset, it is unlikely that further increases in rooftop solar capacity will further reducing peak demand.
Third, demand was much more “peaky” in 2016-17 than in 2007-08. In other words, grid generation capacity needed to meet peak demand was needed less often, meaning it was less efficient to maintain it year round to meet those peaks.
In 2007-08 the top 2% of annual demand occurred during 6 hours over the year, the top 5% over 15 hours and the top 10% during 43 hours. The corresponding figures for 2016-17 were 2 hours, 8 hours, and 16 hours.
Fourth, the minimum annual 30 minute demand (ignoring demand during blackout events) was also lower in 2016-17 than in 2007-08.
This was 661 MW (on 2 October) compared with 996 MW (on 20 April 2008). Furthermore, the minimum occurred during the middle of the day, rather than in the early hours of the morning.
These changes make both demand response and batteries potentially very economically attractive in South Australia. With peak demand events occurring for fewer hours, it becomes increasingly feasible and efficient to meet those peaks through incentives to balance supply and demand by reducing demand for short periods, rather than maintaining supply capacity year round sufficient to meet those peaks.
Alternatively, batteries would only need to be able to deliver output for brief periods to achieve the same outcome.
While rooftop solar has driven changing patterns of demand, wind generation has produced even more substantial changes in ‘residual’ demand – grid demand not met by wind generation.
Figure 14 shows how the residual load duration curve for South Australia changed over a nine year period from 2007-08 to 2016-17.
The substantial gap between the curves represents the large increase in wind generation in South Australia over this period.
Residual demand has also become substantially more ‘peaky’, with similar a similar annual maximum, but the much steeper slope, meaning that the top 10 to 15% of demand occurs for only a few hours in the year.. Median demand left unmet by wind felling by more than a third and is around a third of the peak.
There were much smaller periods in which non-wind grid generating capacity is required to meet the peak and all capacity was much less likely to be operating at any period. The bottom 20% of intervals all had lower residual demand than the minimum demand a year earlier.
In short, increased generation from variable sources not only reduces residual demand but shifts the profile, requiring complementing generation that must operate for fewer hours each year. The graph shows how increased variable wind generation makes battery storage and demand response even more economically attractive – and makes inflexible generation like coal power much less viable.
High renewables, low prices, stable supply: implications for re-designing the National Electricity Market
While the increase in wind power in South Australia has been blamed for blackouts and high prices, the recent experience in South Australia does not support these conclusions.
High levels of variable wind power, often resulting in exports to Victoria, have been complemented by gas generation, resulting in low prices and stable supply.
The recent experience of South Australia has a number of implications for the design of any mechanism to support reliability and emissions reductions, such as the proposed National Energy Guarantee (NEG) scheme.
While the NEG is currently lacking detail, it was released with modelling in which renewable generation at 28-36% in 2030 – with little or no new renewables from the end of the LRET in 2020 – and is touted as supporting ‘dispatchable’ coal fired power plants.
The experience of South Australia in recent months gives grounds for scepticism about such a proposal.
First, average wholesale prices in South Australia throughout the period were lower than in all three other mainland states and only very marginally higher than in Tasmania.
As can be seen below, the two states with the highest prices during the month were Queensland and New South Wales, the two states with the highest shares of coal generation. Limiting new wind capacity and keeping old coal-fired power stations open is not necessary to reduce wholesale prices.
Second, although the share of variable large scale renewable generation was above 50% during September, and just below during October, there were no problems with the reliability of electricity supply.
Moreover, this share was lower than it would have been had AEMO not intervened in the market to curtail wind generation at less than 1,200 MW for a number of hours on various occasions during the month, totalling the equivalent of more than two days.
AEMO’s interventions were taken to guarantee system security. It is therefore reasonable to assume that AEMO considered the electricity supply system in South Australia to be reliable at all times during September and October, despite the high share of variable renewable generation.
During one 30 minute trading interval the wind share of generation was over 80% and it was over 75% during more than one hundred trading intervals in September, summing to the equivalent in total to over four days in the month.
If this is possible in South Australia why should other states be different? If they are not different, then why does the government propose to constrain renewable generation to 36% or less by 2030?
If there is no technical need for such a constraint to guarantee reliability, it is hard to avoid the conclusion that the limitation is driven by ideology.
Third, most of the gas generation during September was supplied by the state’s two combined cycle power stations (and for that reason the two most efficient) – Pelican Point and Osborne. The average capacity factor for the month was 78% at Pelican Point and 76% at Osborne.
Most of the remaining gas generation came from the ageing Torrens Island A and Torrens Island B stations. Adjusting for the fact that two units at Torrens Island B and one at Torrens Island A were shut, presumably for maintenance which is commonly undertaken in spring and autumn, these two plants had capacity factors of 13% and 43% respectively.
The two units in operation at Torrens Island B spent much of the month generating at either 90-95% or at 25-30% of rated capacity. Torrens Island A was totally shut down for six separate extended periods totalling more than half the hours in the month, but was also required to generate at near maximum output over about 6 hours at various times.
Such highly variable modes of operation are necessary for dispatchable generators being tasked to balance supply from variable renewable generators.
Rapid variation in output and operation at low part-loads are technically feasible for gas generators, though newer gas turbine plants are better suited than the ageing gas-fired steam turbines at Torrens Island.
Low capacity operation is also financially feasible, though perhaps not attractive, for old plants like Torrens Island, which are presumably fully depreciated and do not have to carry financing costs.
Elsewhere in the NEM, the rhetoric about the NEG appears to envisage that the variable functionality provided by Torrens Island in South Australia will be provided by coal-fired power stations.
They are not well suited to this role. Technically, coal-fired units cannot operate for extended periods at part loads of less than 50%. Further, they cannot be entirely shut and then restarted over periods of a few days, at least not regularly.
Financially, some may be able to operate for extended periods at low overall capacity factors. If they did, however, some may have difficulty justifying more than routine maintenance costs, undoubtedly induced by duty cycles requiring rapid and large amplitude output variation.
It is presumably such considerations which have led many critics of the NEG to suggest that it will lock in coal-fired generation at high capacity and block the transition of the Australian electricity supply system to a low emission future.
Fourth, by removing demand from the middle of the day, the steadily increasing capacity of rooftop solar generation will also increase the variability of the residual demand which dispatchable generators currently supply.
The designers of the NEG appear to think that it will greatly constrain increases in wind and grid-connected solar generation after the LRET is built out in 2020.
Such a limitation would serve to “protect” the role of existing coal fired power stations, by preventing further increases in the variability of the demand which dispatchable generators are required to meet.
However, the NEG will not, without imposing some form of outright ban, be able to constrain the continuing growth of rooftop solar generation, which will continue to be an attractive investment for many electricity consumers, even without the rebate currently available through the Small Renewable Energy Scheme.
As a final comment, for long-time observers of Australian energy and greenhouse policy there is painful irony, and not a little apprehension in the facts that:
- the NEG will incorporate environmental objectives relating to greenhouse gas emissions into the National Electricity Market Rules;
- in doing so, numerous far-reaching changes to the Rules will have to be completed in two years or less, and
- according to reports, the principal architect of the NEG is the Australian Energy Market Commission (AEMC).
For well over a decade, many individuals and organisations have tried to change the National Electricity Objective (NEO) to incorporate environmental objectives. The NEO is a statement, endorsed by COAG and embodied in the National Electricity Law, which determines the whole framework for the NEM Rules.
To paraphrase, the objective is to promote efficiency in all aspects of the electricity industry operation, so as to enhance, in the long term interests of consumers, price, quality, safety, reliability, and security of supply of electricity.
All attempts to add an environmental objective to this list, and therefore to bring the pursuit of environmental objectives into the scope of the Rules, have been vigorously and effectively opposed by the AEMC.
The AEMC has also, in its role as “the rule maker for Australian electricity and gas markets” become notorious for its slow processes to make any change to the existing Rules. I
It is hard to avoid the conclusion that, faced with an unavoidable need to make some changes in pursuit of emissions reduction, the AEMC has skilfully ensured that it will take the lead in determining how such changes are made and what the new Rules will be.
Hugh Saddler is an energy analyst who compiles the National Energy Emissions Audit on behalf of the Australia Institute’s Climate & Energy program.