The boundary effect

Fundamental to running any power grid – from micro to nationwide – are multiple mechanisms ensuring that one simple equation is continuously satisfied:

Supply = Demand

WattClarity is replete with detailed articles describing these mechanisms and how they’re implemented in the NEM. Amongst these posts are deep dives into how “Supply” and “Demand” are actually measured – the answer comes in many different flavours, using many different inputs, depending on the purpose of the measurement.

This post covers one particular purpose: constructing the demand “snapshot” fed into the NEM’s market scheduling process at the beginning of every five-minute dispatch interval. It concentrates on a specific input to that snapshot concerning the NEM’s inter-regional interconnectors, and then on only one interconnector – Basslink, the direct-current cable linking Victoria and Tasmania. Even more specifically, the focus is on a single issue – where on Basslink does the boundary between the Victorian and Tasmanian regions lie?

This may seem a ridiculously arcane question. But the 1 July 2026 change in Basslink’s status from “market network service provider” (MNSP) to “Regulated interconnector” involves a subtle change in the answer, affecting how the market demand snapshot is measured and calculated. Market data since the changeover date strongly suggests that market operator AEMO didn’t quite nail every required system change, and that there are measurable effects – fortunately none too dramatic – on operation of the Tasmanian power grid.

Because it’s an interesting example of how the grid copes with things like measurement errors leading to frequency drift, as well as illustrating how subtle errors can arise, we’ve written it up here.

We measure Supply – not Demand

The Supply = Demand equation places stringent requirements on operating a power system, but it hugely simplifies demand measurement – we only have to measure Supply and voilà, we have simultaneously measured Demand.

For market dispatch, the flavour of Supply (and hence Demand) that we’re concerned with has just two basic ingredients:

• Output from system’s large-scale “scheduled” and “semi-scheduled” generators (including grid-scale battery discharge)
• Transfers between NEM regions across interconnectors

We can picture how these ingredients are combined to measure regional supply and demand like this:

Each NEM region has a “Regional Reference Node” (RRN), which is both a notional accounting concept, but also a very specific network location on the region’s high voltage grid.

For supply accounting purposes, every generator in the region is considered “virtually connected” directly to that RRN, and generator supply “at the RRN” is the sum of metered output from all generators within the region. There is no adjustment for the real transmission losses between generators and the physical RRN. If you like, we are treating those losses as included in the supply and demand equation in each region.

Losing on transfers

But the equally real transmission losses on power transfers between regions are treated quite differently. The measurement process starts with metered transfers at the regional boundary, but then adjusts these volumes for allocated transfer (“tx”) losses between the boundary and each regional reference node.

In the example above, the NEM’s expressions for supply to (and demand at) each regional reference node are as follows:

Region A supply (demand D_A) = ∑G_i – T_A-B – Region A tx losses

Region B supply (demand D_B) = ∑H_j + T_A-B – Region B tx losses

That is, add up all the generation output, add or subtract the transfer between regions, and adjust for inter-regional losses on that transfer.

The only difference between these expressions, apart from the generator labels, is the sign on the boundary transfer T_A-B. Each NEM interconnector has a sign convention specifying which of the two RRNs it connects is exporting when the flow is positive. In the example, Region A is exporting to Region B when the boundary flow is positive.

But allocated tx losses are always subtracted regardless of flow direction and sign convention – in the case of Region B above, assuming boundary flow T_A-B is positive so B is the importing region, this seems reasonable: imports reaching Region B’s physical RRN will be less than the boundary-metered value because of those losses.

It’s slightly less obvious that to arrive at supply (demand) for Region A we should also subtract allocated losses. What’s happening is that to measure supply required for Region A itself, we not only deduct energy transferred out of Region A at the boundary, but also the component of supply in Region A which is covering losses from the physical RRN to the boundary, which arise only because of that transfer.

By playing with examples, you should also be able to convince yourself that these expressions also work with no changes in sign when the transfer is from Region B into Region A, ie when T_A-B is negative.

Many a slip

So, with that not-too-difficult accounting for supply and demand, how did AEMO slip up with Basslink? The reason is a subtle difference in the treatment of interconnectors depending on whether they are MNSPs or Regulated.

Unlike a Regulated interconnector, an MNSP is an active participant in the energy market. It offers a price to transfer a given maximum volume of power from one region to another (actually, up to 10 distinct prices for each of 10 successive volume bands), and that price is taken into account in the NEM’s economic dispatch process – unless the difference in regional spot prices exceeds the MNSP’s offered transfer price, transfers will not be scheduled. (Note this is now a purely theoretical concept, as Basslink was the last MNSP standing in the NEM after others like Murraylink gave up the fight long ago and transferred to Regulated status.)

The subtlety here is that transfer volumes for MNSP links are specified and measured at their physically receiving (downstream) end, not at a fixed inter-regional boundary point.

This has two specific effects on demand measurement when an MNSP is involved (ie Basslink before 1 July this year):

• The “boundary metering point” swaps between the two ends of the link depending on transfer direction
• All energy losses on the link are allocated to the exporting region, because the link is always entirely upstream of the boundary so effectively located in the exporting region

The next two diagrams show how MNSP link flows and losses are (or were) handled in demand measurement, with the point of transfer – and in effect the regional boundary – being the MNSP’s receiving end. (Excluded from the diagrams are any allocations for losses between the MNSP’s physical terminals and the corresponding RRN locations. These losses are relatively small for Basslink, in fact zero within Tasmania, because the Tasmanian terminal of Basslink is at George Town, which is also the Tasmanian RRN).

When the MNSP flows from Region A to Region B:

And when it flows the other way:

The demand expressions now change depending on the direction of MNSP flow. And while the boundary transfer T_A-B looks identical in both cases, it’s measured at opposite ends of the MNSP link depending on flow direction.

Up until 1 July 2026, AEMO’s systems correctly dealt with this treatment of MNSP boundary metering and loss allocations – analysis of market data clearly shows that all of Basslink’s MNSP losses were being allocated to the exporting region, and that the Tasmanian and Victorian regional demand snapshots used to dispatch the market were being calculated as shown in these diagrams.

Farewell MNSP, hello Regulated Basslink

With its shift in status to Regulated, it became necessary to change the treatment of Basslink losses and the transfer metering point to align with the method used for other Regulated interconnectors:

Regardless of flow direction, the boundary between Victoria and Tasmania is now fixed at the George Town RRN in Tasmania (how many Tasmanians know about this border incursion I wonder ?). With Basslink in effect now entirely within the Victorian region, all losses on transfers get allocated to Victoria. The supply and demand expressions no longer change depending on the direction of Basslink flow.

However, analysing AEMO’s market data from 1 July 2026 now shows that:

• Basslink losses are always allocated to Victoria – correct, but
• Regional supply is likely still being calculated using transfer volumes (T_A-B) measured at the receiving end of Basslink – incorrect when Basslink flows from Tasmania to Victoria

The evidence, Your Honour

We can illustrate the latter problem by comparing AEMO’s published “INITIALSUPPLY” figure for Tasmania – which should correspond to the Region A supply expression above – with a bottom-up calculation using the Tasmanian unit-level generation outputs and metered Basslink flows published in AEMO’s MMS dataset.

Since 1 July, Tasmanian INITIALSUPPLY has been overstated by about 15-30 MW when Basslink flows moderately to strongly northward. This difference corresponds quite closely with the expected link losses across Basslink at these flow levels. There is no discrepancy when Basslink flows southward to Tasmania.

AEMO’s INITIALSUPPLY data for Victoria show exactly offsetting discrepancies to Tasmania’s since 1 July 2026.

This is very strong evidence that when Basslink flows north, AEMO’s INITIALSUPPLY calculation is still being fed internally from metering at the Victorian terminal of Basslink, as if Basslink were still an MNSP, and not from the George Town boundary metered flow also published by AEMO. The difference between metered flows at the two ends of Basslink is of course the link’s actual loss level. The diagram below shows what seems to be going on and why it’s causing these discrepancies:

Not just academic

All this might seem only of interest to NEM demand tragics, but it does have observable impacts on the power system in Tasmania.

Because AEMO’s market dispatch system is being fed with too-high a figure for actual on-island Tasmanian demand when Basslink exports, that system is then producing five-minute generation targets that are slightly too high for the likely demand level at the end of each interval.

The effect of this error could show up in a few ways:

• Tasmanian system frequency may start to rise above the target level of 50 Hz, which is the natural response of an AC grid to supply beginning to exceed demand,
• In response to rising frequency, generators’ primary frequency response capability will act to reduce output in proportion to the frequency deviation, and
• AEMO’s AGC system may also start sending signals to certain Tasmanian generators to further reduce output, based on their enablement for Lower Regulation FCAS, a centralised mechanism for helping steer system frequency back towards the target 50 Hz.

The final chart illustrates that the last two effects can be seen in Tasmania since 1 July, when Basslink is exporting to Victoria.

The chart plots aggregate Tasmanian generator deviations from their output targets, and utilisation of Regulation FCAS capability, for all dispatch intervals since 1 July 2025.

Generator deviations include all differences between scheduled targets and actual end of interval output due to primary frequency response, utilisation of Raise & Lower Regulation FCAS (RREG and LREG), as well as other sources of error. The chart shows that since 1 July 2026 these deviations are clearly skewed downwards – ie outputs fall below target – at higher levels of Basslink export to Victoria. There are no corresponding trends on either side of 1 July 2026 when Basslink flows into Tasmania.

Additionally, the chart shows that since 1 July 2026, again when Basslink is exporting to Victoria, Lower Regulation FCAS is more highly utilised – expected in response to rising frequency – than is Raise Regulation, and that this utilisation also skews noticeably with higher levels of export.

Prior to Basslink becoming Regulated, no similar trends were obvious, again pointing to an asymmetric problem with AEMO’s demand calculation following that change in status.

Fortunately this should be easy enough for AEMO to fix: simply by pointing the demand snapshot calculation at the correct source of Basslink metered flow data, the George Town end of the link. And with no MNSPs now left in the NEM, it’s an error that’s unlikely to recur anytime soon.

 

=================================================================================================

About our Guest Author