This fourth and final article in our four-part series, following on from part 1, part 2 and part 3, asks whether MPFR and FPP-Reg-FCAS Recovery have succeeded, and looks at what the data suggests for the future of Regulation FCAS procurement and AGC.

Did MPFR succeed?

The MPFR reform set out to reintroduce the primary control technical layer inside the Normal Operating Frequency Band. On the evidence discussed in Article 1, including Linton’s long-range frequency charts and AEMO’s current frequency monitoring figures, that reform has succeeded. MPFR has been implemented across scheduled generating units and batteries, and frequency control has been restored inside the NOFB.

That does not mean dispatch errors have disappeared. MPFR provides the fast local response that keeps frequency under control, but secondary control through AGC and Regulation FCAS is still needed to correct errors that persist, relieve the work being done by primary controllers, and restore energy balance.

Did FPP- Reg-FCAS Recovery succeed?

FPP-Reg-FCAS Recovery aimed to quantify and price dispatch errors so units that cause errors pay units that help correct them. Article 2 explained the mechanics: FPP identifies good and bad deviations, measures corrective work through RCR, and converts that performance into settlement outcomes. Article 3 then showed that the resulting settlement patterns broadly align with the reform design, with controllable scheduled plant more often receiving credits and forecast or resource-driven errors more often facing charges.

On that basis, FPP-Reg-FCAS Recovery has succeeded in creating a settlement framework for dispatch errors. Dispatch errors are now measured and settled in a way that appears consistent with the design intent. At the same time, because the mechanism uses the co-optimised Regulation FCAS market as an input, it also inherits any imperfections in how that market is procured and used.

Has behaviour changed?

The settlement patterns, scale of dispatch errors, and the extent to which AGC-Regulation FCAS units correct them, provide useful insights for both market participants and AEMO. Participants can consider the cost of dispatch errors when pricing energy and FCAS and when deciding whether unit control is more profitable than generating at the margin. The range in settlements between Semi-Scheduled units does indicate that participants can manage these costs if profitable to do so, and this will vary on energy, PPA LGC prices, and security constraint curtailment which affect the profitability of the unit at the time.

What comes next?

FPP data may also highlight opportunities to improve how AEMO procures Regulation FCAS and how AGC uses enabled units. This article suggests that measurements of the net error presented by non-scheduled load, residual forecast error and semi-scheduled units could help inform future Regulation FCAS requirements, because those errors appear to drive FPP-Reg-FCAS Recovery settlement patterns.

AEMO has itself outlined priority actions that include trialling AGC changes and more dynamic procurement of Regulation FCAS. Those actions seem sensible, and ideally, they would be informed by, or supported by, FPP data.

FPP-Reg-FCAS Recovery may allow for dynamic procurement of Regulation FCAS

The next chart uses 5-minute data from AEMO’s FPP database to compare RCR, Regulation usage converted to MW, and Regulation enablement. Although it covers only a single day, the data shows clear periods where RCR rises, Regulation-enabled units are used heavily in one direction, and the opposing service is used lightly. Despite this, the dispatched Regulation FCAS quantities remain relatively static because AEMO’s requirements are not dynamically adjusted to reflect changing system conditions.

 

The quarterly averages show that dispatch errors persist at certain times of day and in certain directions.

This suggests Regulation FCAS requirements could be adjusted more dynamically. For example, they could be increased at certain times of day where more variation is expected, or requirements could shift from LOWERREG to RAISEREG at times, and in the opposite direction at others. To the extent that FPP-Reg-FCAS Recovery uses the FCAS Regulation market as its input — as it should, because Regulation FCAS is a co-optimised service expressly designed to correct dispatch errors — it adopts any imperfections of that market.

If system dispatch errors are high and usage of Regulation FCAS is high, that indicates demand for secondary control. Incremental procurement of Regulation FCAS could be factored automatically into the constraints, the price would go up, and units on AGC-Regulation would capture greater FPP revenues and may reduce primary control response. Conversely, if errors are low and usage is low, the opposite could occur.

As previously identified, RCR is not a good candidate to do drive Regulation FCAS procurement, as it is a gross dispatch error whose aim is to ensure sufficient settlements in FPP. The FM is not a good candidate either, because although it can be converted into a MW value, this requires AGC to estimate the primary response on the system (explainer later in this article) and is subject to error.

Could the FPP patterns settlements also drive procurement of Regulation FCAS?

The previous articles identified the broad settlement pattern from forecast elements (residual), solar and wind farms to scheduled plant like batteries.  There was clearly random offsetting behaviour between the variable renewable units despite the cohort generally being identified a bad under FPP.  Isolating this random offsetting behaviour from units that are not providing MPFR or AGC response may be used to estimate the underlying dispatch error that Regulation FCAS needs to correct.

For example, Regulation FCAS could be targeted at the net dispatch error from semi-scheduled units and the residual category. Put another way, the requirement could be derived from the inverse of scheduled-unit deviations, which are typically classified as good performance. The point is that if the net dispatch error from semi-scheduled units and the residual category is driving FPP-Reg-FCAS Recovery outcomes then maybe it should also drive Regulation FCAS outcomes.  Recent interval data could then inform FCAS requirements for upcoming intervals. If dispatch errors and Regulation usage are both high, just as traders may consider this when deciding whether to supply FCAS, AEMO could dynamically procure more Regulation FCAS. This would likely increase Regulation FCAS prices, Reg-FCAS Recovery amounts and FPP settlements, thereby raising the marginal price signal for dispatch errors.

The following chart compares net semi-scheduled deviations with and without the residual deviation (with is “net_err_NS”), Regulation usage and Regulation requirements. During this period, the enabled raise quantities are used quite heavily by AGC, while the enabled lower quantities are used little. Consistent with charts from previous articles, the Regulation and Usage MW are only shown when they align with the Frequency Measure, allowing comparison with FPP’s assessment of their performance.

Aside from this period indicating that maybe more Regulation FCAS Raise could have been enabled, there’s a quite consistent relationship between the FM, Scheduled units and Regulation Usage, which is good. Yet there are periods where the FM seems to switch more readily than the net errors (maybe a function of the tight frequency deadband of 15mHz under MPFR). An initial hypothesis would be the dispatch error causing high frequency (FM being negative) is managed by units with positive errors, reducing that positive error, but remaining positive overall, and then being classified by a FM that has now crossed 50Hz as being bad.

The FM was chosen for good reasons: it gave plant controllers confidence that units applying compliant MPFR settings would be assessed as “good” under FPP-Reg-FCAS Recovery. That remains a sound foundation. However, MPFR is now embedded in the Rules, and industry better understands the importance of a technical layer of control within the normal operating band. With frequency tightly controlled around 50 Hz and a narrow 15 MHz deadband, frequency can cross the target frequently.

Looking again at the chart for Regulation enabled units, it is worth observing the Raise regulation enabled units are always positive, and for periods where there is no shaded area (i.e. not green, not shaded at all, see red arrow), then these positive errors are classed as bad in FPP.

These raise units that may have increased dispatch in response to AGC, may largely accommodate the need to reduce dispatch error by reducing generation as their AGC raise signals fall away to the scheduled trajectory, or simply the trajectory catching up with their generation, or some limited primary control response, or a mix of all three – given there is crossover between raise and lower enabled units, it may be the case that they are signalled by AGC to reduce dispatch, but this is unlikely given AGC has set rules, and max/min limits as to when AGC signals can be sent to the unit – and a unit that has responded to raise AGC signals and is above the trajectory may not be within these limits to receive lower AGC signals. Without more information on a unit’s AGC model and the AGC SCADA signals themselves, it is difficult to ascertain – however what can be understood is the change in dispatch error and resulting FM was largely accommodated by those Regulation Raise units reducing their positive dispatch error, but remaining positive overall, and therefore at times being categorised as bad by the FM.

Longer timeseries

Now zooming out with a longer timeseries let’s compare measures of system error, such as isolating random offsetting behaviour from semi-scheduled plant, the residual error and comparing with the scheduled units that provide primary and secondary control.

Thinking of Raise and Lower as ranges of likely errors, the RCR values widen noticeably during the day, as do the Usage values, however this is biased positively (raise) as shown in yellow.

To compare with RCR and Usage which are the greatest of the 75 4-sec values during a dispatch interval, the maximum and minimum net deviations of scheduling cohorts is plotted.  Clear is a change in the daytime, where the net error of the semi-scheduled units (shown as a range in green) dips and widens during the day.  The red lines are semi-scheduled units coupled with other non-scheduled elements including the residual. The scheduled units, shown as a range in blue (the inverse of the non-scheduled, semi-scheduled and residual as shown in red) is positive and shifts upwards, showing the prevailing need for raise regulation FCAS correction, and not lower during the daytime.

These charts reflect the settlement dynamic of FPP and Reg-FCAS Recovery. FPP prices all errors by targeting the gross error – and creates a general payment from non-scheduled elements (think consumer load) and semi-scheduled units to scheduled one, think batteries and Regulation units in particular – but FPP only pays for what is used in real time. By contrast Regulation FCAS is an estimate of the net error and is paid for irrespective of whether it is used.

If we look at the above charts, given the prevailing errors seem to driven by semi-scheduled units being below their scheduled targets, and this frequently being compounded by the residual, it may be that both Regulation FCAS and FPP create unnecessary settlements for Lower direction errors – in the case of Regulation FCAS by having ex-ante enabled quantities that aren’t much used, and FPP by using an FM that, despite being smoothed and designed to represent good control action, seems to change sign frequently when maybe the errors are being accommodated by reducing a positive dispatch errors on a scheduled unit.  This seems to be supported by the data, shown in the chart below, where total good errors in the lower direction are of the same scale as total good errors in the raise direction.

While Article 2 identified that the Lower prices and settlements are far less material than Raise, an equal balancing of good deviations in each direction, as determined by the FM and shown in the chart above, is unsupported by the hypothesis that errors are placed on the system by semi-scheduled and forecast elements. That hypothesis is supported by the general financial transfers in Article 3. The fact that the settlement patterns better reflect the errors of these cohorts and their known control capabilities is more comforting, but it suggests there is scope to better understand the merits of judging performance using the FM when operating with the very tight 15 mHz deadband for MPFR.

Can AGC use Regulation FCAS more effectively?

More dynamic FCAS requirements may improve how efficiently Regulation FCAS is procured. Yet buying more Regulation FCAS will not be effective if AGC won’t use it.

AGC determines the required adjustments from enabled units using proportional and integral calculations of area control error. Area control error estimates the system’s MW imbalance for a given frequency deviation, using an empirical MW/Hz relationship between frequency and energy changes. The proportional component converts the frequency error into an immediate MW response, like a primary control signal. The integral component accumulates that response over time. AGC then combines, smooths and filters these signals through deadbands, range limits, before allocating adjustments to Regulation FCAS units.

However, when many generators and batteries are operating under MPFR, frequency may become less sensitive to dispatch errors, or the actual MW/Hz relationship may differ from the estimate used by AGC. In those conditions, AGC may underuse Regulation FCAS units. Retuning AGC or adapting it to respond more directly to the underlying dispatch error, could allow AGC-Regulation FCAS to correct errors faster and reduce primary response, and increase FPP settlements for FCAS-enabled units. One option would be to rely less on proportional and integral signals that are based on frequency and instead use 4-second measurements to estimate system error. This error could then be allocated directly to Regulation FCAS units.  AGC itself has such a feature, called basepoint adjustment, which AEMO describes as:

“Basepoint adjustment in AGC is an additional calculation that can reduce deviations from energy target on generation units providing PFR, and transfer that generation to units providing regulation FCAS. AEMO considers it may offer a practical means to increase the role of regulation FCAS.” – AEMO, Technical Review of the NEM Frequency Control Landscape

It would be sensible to support AEMO colleagues as they trial any improvements in dispatch, AGC and Regulation FCAS and FPP.

 

---

About our Guest Author

---