The NEM – Part 6: Ancillary Services
The NEM's favourite four letter word begins with an F
A reminder that we’re hosting a trivia event in conjunction with Australian Energy Week. The trivia night will be 6 pm Tuesday 9th June, at The Decks at The Boatbuilders Yard, just outside the Melbourne Convention and Exhibition Centre where the conference is being held.
Tickets are available here: https://events.humanitix.com/aew-currently-speaking-trivia. In the past these events have sold out, so don’t wait to get your ticket!
We’re excited to be working Australian Energy Week — this year’s event has some huge names from across all sectors of the industry, as well as specialised pre-conference masterclasses for participants looking to up-skill in some of the niches and vagaries of energy markets.
It’s a must attend event if you’re looking to stay across an industry which is rapidly evolving.
Final update — a huge thank you to all of the early stage career professionals who submitted an entry for an AEW conference ticket. We’ve closed the form and are in the process of reviewing the entries. We’ll be in contact with the winners via email shortly!
This is a continuation of the NEM Explainer Series.
If you’re new here, we suggest starting at the Overview of the NEM.
Baby don’t Hertz me
The NEM, like all major electricity supply systems, operates on Alternating Current (AC).
Alternating Current supplies electricity in a sine wave shape — the voltage of the electricity delivered varies over time, flipping signage (reversing polarity) between positive and negative.
If we measure the time taken for the pattern to repeat, we have the period of the wave, and if we take the inverse of the period we have the frequency, measured in Hertz. The frequency is simply a measure of the number of cycles per second.
This is shown visually in the diagram below – a 50 Hz wave has a period of 20 milliseconds (i.e. 50 Hz = 1 / 0.02 seconds).
After experimenting with different frequencies in the first half of the twentieth century1, all major electricity systems have landed on standardised frequencies of either 50 or 60 Hz. 50 Hz is far more common globally; only the United States, half of Japan and some other parts of Asia utilise 60 Hz.
Delivering power at this constant system frequency is extremely important — both generators and devices are designed to operate within a relatively tight frequency range (one or two percent) around the nominal frequency.
Beyond being standardised, an important property of the grid frequency is that it provides a measure of the supply and demand balance – more supply (generation) than demand will push the frequency above the nominal operating frequency, and vice versa.
In practice supply and demand, no matter how hard we try, are never perfectly matched, and so the system frequency wanders around a little bit.
Power systems procure ancillary services via contracts or markets in order to incentivise generators (and loads) to actively manage the system frequency to ensure stability. Ancillary services can also be used to manage other power system characteristics like voltage or oscillatory stability or other increasingly esoteric (but important!) attributes; however ancillary services for frequency control are a near universal feature of all modern electrical grids.
A somewhat crude analogy regularly invoked is to think of the system frequency as the speed of the system. Take your foot off the accelerator pedal (too little generation) or drive up an incline (too much demand) and the speed will drop. This requires correction lest the grid (somewhat literally) stalls.
Enter FCAS
The NEM has a marketised2 approach to procuring these ancillary services for frequency control — literally called the Frequency Control Ancillary Services (FCAS) – across 10 separate markets. I don’t want to say 10 individual markets are overkill, but the NEM has more frequency control ancillary services than any other grid I’ve encountered; most grids achieve the same outcome with half a dozen or less services. 🤷♂️
There are two Regulation FCAS markets – Raise Regulation and Lower Regulation FCAS. The purpose of these markets is to maintain the system frequency within tight operating limits during normal operation. This operating envelope is known as the Normal Operating Frequency Band (NOFB), and on the mainland is 49.85–50.15 Hz (±0.3% of the nominal 50 Hz).
If we return to our speed analogy, the simplest way to think of the regulation FCAS markets is that they’re responsible for constantly modulating the grid’s accelerator pedal to maintain a consistent speed. Generator (or loads) constantly make small changes to their output (up or down) in order to keep the frequency within the NOFB.
The remaining eight markets are split between the Raise Contingency and Lower Contingency services. These services exist to manage frequency deviations when something unexpected happens — known as a contingency event.
Both the raise and lower sides are split into the Very Fast, Fast, Slow and Delayed markets. These markets are named according to the speed of response required:
Very Fast Raise/Lower Contingency FCAS requires a response within 1 second, sustaining that response for 6 seconds.
Fast Raise/Lower Contingency FCAS requires a response within 6 seconds, sustaining that response for 60 seconds.
Slow Raise/Lower Contingency FCAS requires a response within 60 seconds, sustaining that response for up to 5 minutes.
Delayed Raise/Lower Contingency FCAS requires a response within 5 minutes, sustaining that response for up to 10 minutes (or less, if the system frequency has been restored to within the NOFB in less than 10 minutes).
These markets are also often referred to via their timings – e.g. RAISE1SEC/LOWER1SEC or R1/L1, RAISE6SEC/LOWER6SEC or R6/L6, RAISE60SEC/LOWER60SEC or R60/L60 and RAISE5MIN/LOWER5MIN or R5/L5.
And yes, it bothers me to no end that the delayed markets are not designated as R300/L300.
The other important naming convention across both the Regulation and Contingency FCAS markets is that the Raise/Lower nomenclature refers to the actions a generator would take to correct the frequency, i.e. Raise = increased generation (or decreased load) = frequency below 50 Hz.
If you’re a visual learner, here is a chart I crudely drew in 2019 and I will continue to produce unsolicited on any occasion that I reasonably can until someone shouts at me… and then I’ll show it a couple more times for good measure.
To recap – the Regulation services exist to keep the frequency within an acceptable range, and beyond that the Contingency services are there to wrangle the system frequency back into the green.
If the system frequency strays too far (and/or too quickly), we end up in the Not good things zone, which likely involves under-frequency load shedding (UFLS)3 where blocks of load across distributed zone substations are automatically disconnected from the grid (blacking out thousands of customers and removing load) in order to prevent wide-scale collapse of the grid.
If the changes via FCAS and UFLS are insufficient to arrest a frequency collapse and the frequency strays into the Really bad things zone, then very large numbers of customers will be blacked out, if not an entire state.
Finally, the original 8 FCAS markets were introduced to the NEM in 2001 (they were contracted for the first few years of the NEM), and in 2021 the Very Fast Raise/Lower Contingency FCAS markets were introduced (beginning in October 2023).
What is a ‘contingency event’?
The most common form of a contingency event is the loss of a large power station, something which happens much more frequently then you might expect. Oh look — here’s a recent one.4
When this occurs – say one of the large coal-fired generator units between 350 and 750 MW in nameplate trips offline suddenly – there will be a corresponding drop in frequency. If the frequency falls below 49.85 Hz (outside of the NOFB), then units providing Contingency FCAS will be expected to respond by injecting active power into the system.5
The chart below (sourced from an excellent WattClarity FCAS explainer) shows how a frequency drop plays out, and the role of the different Contingency FCAS markets in arresting, stabilising and restoring the system frequency back to 50 Hz.
On the flip side, the loss of a large load (in practice often a network element carrying a large chunk of load) will cause a corresponding increasing in frequency above 50 Hz, and the same things will occur in the Lower Contingency markets.
Historically these over-frequency events occur significantly less than under-frequency events and the NEM is somewhat unusual in even having lower contingency frequency ancillary services.6
Frequency is a ‘global’ property
Electricity (well, current) doesn’t quite travel at the speed of light, but it’s not far off. So if the system is fully interconnected, the frequency is roughly the same across the entire grid, from Port Douglas to Port Lincoln.7
This has three implications:
Any generator can inject (or remove) power anywhere in the system in order to correct the cause of a frequency disturbance anywhere else.
The frequency at the point of the generator providing the response will be (almost) the same as the point of the frequency disturbance. This means that the generator providing the frequency response can rely on the localised frequency measurement, rather than a centralised dispatch signal.
If frequency disturbances anywhere in the grid can be managed by any generator located anywhere else in the grid, the price of each FCAS market must be the same across all regions.
Point 3 is a funny one, because the FCAS prices, like the energy spot price, are regional. This means there are 50 different FCAS market prices – one for each of the 10 markets in each of the 5 regions. Silly talk.
Sillier still when the prices are the same across all regions for something like >95% of the time.8
Except when it’s not
When the system is not fully interconnected – i.e. when one of the regions is fully or partially islanded – or when the system is at a credible risk of separation, then additional FCAS quantities will need to be procured within the islanded region.
For example if the Heywood interconnector between Victoria and South Australia is on outage, SA is islanded from the broader NEM and all FCAS quantities must be procured from generators (or loads) within SA.
In this instance the FCAS prices in the other regions will be the ‘global’ price, and the FCAS prices within SA will be the global price plus the local price, which reflects the additional cost of only procuring FCAS from SA generators (or loads).
For hopefully obvious reasons FCAS price separation is most likely in SA, QLD and TAS.
FCAS Pricing and Procurement
This dovetails into how FCAS prices are set.
The NEM Dispatch Engine (NEMDE) determines the level of FCAS required to manage contingencies on the system (based on various constraints).
For the 8 contingency markets, the simple explanation is that there needs to be enough FCAS to cover the loss of the largest generator, after accounting for load relief.
Load relief is the estimated inertia which exists on the demand side – e.g. large synchronous motors and pumps which will change power draw as the grid frequency changes (because a motor’s rpm is proportional to the frequency). This load relief effectively provides an FCAS-like response for free.
A back of the envelope calculation looks something like: Kogan Creek (750 MW) is online at full tilt and the mainland NEM is operating at a total operational demand of 30 GW, then 0.5% (the current estimated load relief figure) of the demand, 150 MW, is provided by load relief, and only 600 MW of Raise Contingency FCAS (across each of the 4 markets) is required to cover the loss of Kogan.
The FCAS markets, like the energy spot market, are dispatched and settled on a 5-minute basis, and centrally dispatched generators (and scheduled loads) submit bids and rebids for each FCAS market alongside their energy bids.9
In fact NEMDE co-optimises the dispatch of FCAS and energy — something which is a world-leading feature of the NEM.
Prices for each market are set in the same way as energy – offers are made across 10 tranches and the marginal bid sets the price. Bids in each price tranche are made in integer MW values.
Where FCAS pricing does differ from the energy market is that FCAS prices are capacity prices — units are paid for their ability or availability to provide FCAS, not for the actual delivered energy. Pricing is thus in $/MW/h, which confusingly is dimensionally the same as energy’s $/MWh.
In fact, this is one of the key misunderstandings relating to the FCAS markets. There is no first order relationship between FCAS prices and the number or severity of contingency events. FCAS prices are a function of the contingency size and load relief calculation (i.e. the amount of procured FCAS), and the bid stack (which has historically been very inelastic).10
Two final points on pricing worth noting:
Average FCAS prices, even during the frothy years circa 2016-2024, are an order of magnitude (or two) lower than average energy prices. The spot market hovers somewhere in the $50-$200/MWh range on average, whereas a monthly average price of $1-5/MW/h for an FCAS market would be good.
Historically there has been a structural price differential between the Lower and Raise Contingency markets. Until recently Contingency FCAS prices were largely set by thermal and hydro-electric stations. For these types of units providing a Raise response incurs an opportunity cost – headroom must be left on the generator in order for more power to pushed into the system. Responding to any actual contingencies also incurs a fuel cost and potentially even an energy spot price below the unit SRMC. Lower Contingency FCAS responses on the other hand simply requiring backing the unit off for a couple of minutes.11
I put the horse before the cart and wrote about some of these structural shifts in FCAS pricing earlier this year.
Technological requirements
Which units can participate in the FCAS markets?
In the case of the Regulation markets, the unit must be centrally dispatched and on Automatic Generation Control (AGC). AGC is a SCADA connection whereby AEMO can measure the unit’s output and provide generation instructions every 4 seconds. In effect this limits participation in the Regulation markets to Scheduled assets – thermal and hydro units, large-scale batteries and yes even wind or solar farms (Semi-Scheduled assets).12
Participation in the Contingency markets on the other hand requires central dispatch (i.e. submitting bids and monitoring dispatch instructions) and specific high speed metering capable of measuring the unit response on a sub-second basis. There also needs to be some kind of local control system capable of measuring the local frequency and triggering a response when the conditions have been met.
Participation in the Contingency markets is thus open to smaller batteries and demand response, including aggregated fleets and virtual power plants — both residential and commercial & industrial.
The technical rules defining all of these are contained within the Market Ancillary Service Specification (MASS).
Cost recovery
The last bit to cover off FCAS 101 is cost recovery.
Contingency FCAS is recovered based on a percentage split of a participant’s usage of the system.
Raise Contingency is recovered from generators – for each 5-minute dispatch interval a participant’s generation output as a percentage of the total regional generation is calculated and then multiplied by the cost of procuring Raise Contingency services in that interval. For example if AGL’s portfolio of assets generated an average 1,000 MW in Victoria in one interval, and the total average Victorian demand for that interval was 5,000 MW, then AGL pays 20% of the cost of procuring Raise Contingency services for that interval.13
Lower Contingency is recovered from loads (customers) via the same concept, which means retailers are responsible for the majority of payments (which they then recover from end residential and business customers).
Regulation FCAS is recovered based on the concept of causer pays. This goes well beyond the scope of an introductory article and has changed recently with the introduction of Frequency Performance Payments (FPP) in 2025. But the (very) high level summary is that generators are expected to ramp linearly between 5-minute dispatch targets — deviations away from this line, in a direction unhelpful to the system frequency (i.e. if frequency is < 50 Hz and a unit is below its dispatch target), will lead to the owners of those units paying more.
The big change with FPP, aka “double-sided causer pays”, is that there is now a stick and carrot approach to these deviations — helpful deviations to system frequency are now rewarded… via a delightfully complicated settlement process.
Non-Market Ancillary Services
Before we wrap, let’s briefly talk about the forgotten neglected cousin of FCAS.
In electricity-markets-that-are-not-the-NEM, it’s not uncommon to see ancillary services, including frequency management-based ones, procured via contract rather than a market.14
In the NEM there a range of other ancillary services, dubbed the non-market ancillary services (NMAS), which are procured via tendered contracts.
The services procured via NMAS have changed over the years, and will continue to change in the near future, but one of the primary services consistently procured is the system restart ancillary service (SRAS) aka black start capability.15
What is black start capability? My favourite (unfun) fact of all time is that power stations require power to run, and more importantly a functioning grid to synchronise to. Some power stations are designed specifically in order to start under their own steam (yes, pun intended) typically via a smaller auxiliary package boiler, which provides power to get the main unit up and running. These units can then begin generating into the grid, without requiring an existing grid frequency to synchronise to.
AEMO has a dedicated page on FCAS’ poorer cousin.
25 Hz was a common frequency utilised in traction power systems (trams and trains), and 40 Hz was also common in mining systems. The frequency of the electricity is directly proportional to speed at which a motor turns (revolutions per minute (rpm) is effectively just alternative measure to Hertz), and so early system designs were dictated by the requirements of the equipment being driven.
Unfun fact — Western Australia had a 40 Hz based system (which was common in early mining operations) for some time, before adopting a 50 Hz standard.
You might think that this is not a word. But if American English has taught us anything, it’s that we can verbify just about anything.
In New Zealand this mechanism is known as automatic under-frequency load shedding — AUFLS — which is a superbly appropriate acronym.
The frequency didn’t breach the NOFB in this instance, but you can clearly see the effect of 700+ MW disappearing suddenly.
Equivalently, a load can be reduced via demand response.
For example New Zealand’s NZEM have the FIR and SIR markets, which are broadly equivalent to Fast and Slow Raise Contingency FCAS. There are no markets for dealing with unexpected frequency rises, likely on the basis that these events are much less common, and can be dealt with via existing synchronous generation.
Correction from Chris Otton: the NZEM has a contracted over frequency arming scheme.
If electromagnetic waves propagate at something like 90% of the speed of light through the transmission lines, then a frequency disturbance in Port Lincoln will manifest in Port Douglas 5,000 km away something like 1 second later (ignoring the dampening effect of inertia in the grid).
I could pull the actual number, but instead I’m going to vibe it.
Also somewhat fascinatingly, all FCAS markets have always been dispatched and settled on a 5-minute basis, long before the 5MS rule change.
There could be an argument for second order effects, where savvy participants structure their FCAS bids such that the price is highest when a greater number of contingency events might be expected, but 1) contingency events by their nature are largely random and 2) there are usually more important bidding strategies at play. This is especially true in the post mandatory primary frequency control world, which has seen an order of magnitude decline in the number of contingency events.
I’ll leave this as a thought experiment for readers to consider how these structural differences might change in a world where the marginal bid is primarily from batteries.
A single large (multi-MW) load, like an aluminium smelter, could theoretically participate in the Regulation markets, but the complexity involved is not insignificant. If anyone is aware of large loads participating in equivalent regulation frequency markets elsewhere in the world I would love to hear about it!
There’s a little more complexity here relating to interconnector flows and some cost sharing between the Delayed and Regulation services. But we’ll gloss over these.
Prior to the WEM market reform and introduction of new FCESS markets, the equivalent regulation and contingency markets were contracted. The NZEM has markets for contingency FCAS equivalents, but regulation (frequency keeping) is contracted.
Correction from Chris Otton: the NZ frequency keeping market is optimised against energy dispatch on a half hourly basis using FK offers provided by generators.
Prior to the WEM market reform and introduction of the FCESS markets, the equivalent contingency service was called SRAS — the Spinning Reserve Ancillary Service. Not confusing at all. The equivalent regulation service was called Load Following Ancillary Service, LFAS, not to be confused with everyone’s favourite forever chemical and power station fire retardant PFAS.
Hey Alex, I think you forgot to finish this example
"For example if AGL generated an average 1,000 MW in Victoria,"
Great work - the right amount of detail and relatable language. Alex, your next diagram could be of the entire system (NEM) and its flows - channel my fav turtle, Michaelangelo.