Semi-dispatch caps are loading the deviation dice

The impact on NEM system frequency of “VRE under-performance” (VRE generators delivering less power than their dispatch targets) is a frequently discussed issue here on WattClarity.

 

I believe it’s become a substantially bigger problem than necessary, because aggregate under-performance is an entirely predictable consequence of the NEM’s dispatch arrangements for semi-scheduled generators, in particular the semi-dispatch cap.

If none of that terminology makes sense, don’t read on. This is a short post aimed at those familiar with these arrangements, outlining the very obvious challenge that semi-dispatch caps create for the NEM’s frequency control mechanisms.

The rules

Briefly, the dispatch compliance rules require that semi-scheduled generators:

• When not subject to a semi-dispatch cap (SDC off): should only deviate from a dispatch target to the extent available power exceeds or falls below that target.
• When subject to a semi-dispatch cap (SDC on): must meet an additional obligation to not exceed the dispatch target.

There are two principal reasons for a semi-scheduled unit’s SDC status to be set on by the dispatch process:

• The unit is subject to a binding constraint which limits unit targets for security or other reasons – generators exceeding targets could result in insecure outcomes; or
• The unit is wholly or partly priced out of the market or marginal to energy supply-demand balance, with its target being set below availability for economic dispatch

When its SDC is off in a dispatch interval, a semi-scheduled generator remains compliant when generating whatever level of available power its energy source allows. So it can deviate both upwards or downwards from target in line with its source energy. But when its SDC is on the rules limit its output to no more than target even if more source energy is available.

Units with SDC off status will necessarily have a target equal to their dispatch availability – being dispatched below availability must reflect one of the two reasons above, meaning the SDC goes on. (Units with SDC on may also have a target equal to availability – for example when in a binding constraint but not curtailed by that constraint.)

Loading the deviation dice

Dispatch availability is a 5 minute ahead forecast. So actual availability and hence power generated will almost always vary from that forecast for VRE units, because their source energy is, well, variable.

This variability interacting with the SDC rules creates an obvious asymmetry: if actual available power at end of dispatch interval for a VRE unit can vary randomly and with more or less equal probability above or below forecast dispatch availability, then units with SDC off will have a roughly equal chance of delivering more or less power than target by the end of the interval.

But from units with SDC on, we should expect only downward deviations, because deviations above target are non-compliant under the semi-dispatch rules.

If semi-dispatch caps apply to a significant proportion of the VRE fleet in any interval, it is almost certain that this subset of the fleet will, in aggregate, under-deliver against dispatch targets, because there will be no units with upward deviations able to offset downward deviations at others. If deviations within the group of non-capped units more or less balance out, this will leave the VRE fleet as a whole with significant net under-performance.

Theory meets practice

To illustrate, here’s aggregate NEM large scale solar performance on 16 January this year, the subject of a note from Paul on frequency performance during the day:

The middle panel shows deviations from dispatch target, grouped into those that were positive at individual VRE units and those that were negative. The negative set tend to be larger so in the bottom panel for the fleet as a whole, we see significant and sustained under-delivery against targets.

But when we further split the data according to SDC status for each unit in each interval, it’s very clear what’s driving the sustained asymmetrical performance:

As predicted, it’s the impact of the semi-dispatch caps preventing almost all positive deviations for that group.

While the uncapped group can still deliver large aggregate deviations in individual intervals, these excursions are shorter and spread between positive and negative. Not so for the capped group, where under-performance is virtually guaranteed across the day.

Not just isolated days

Longer term analysis of these patterns shows consistent results. Averaging across an entire year smooths away large outlier values for particular days and intervals, but retains the expected pattern of smaller aggregate deviations when not subject to semi-dispatch capping, and substantially larger and negative aggregate deviations when capped.

 

Interestingly, uncapped group deviations are positive on average, which might indicate some slight inherent conservatism in dispatch availability forecasts.

Any suggestions?

What are the implications of these results? More consideration is required, but a few thoughts spring to mind:

• Since the semi-dispatch rules virtually guarantee under-delivery, potentially large, from the group of units with SDC on, then would it make sense to enable larger volumes of raise regulation FCAS whenever SDC is on for a significant volume of available VRE, so that frequency drift below 50 Hz is better controlled?
• Are VRE facilities able to comply adequately with the NEM’s Primary Frequency Response (PFR) requirements? If so, are units with SDC on and actual availability above target being inhibited from delivering PFR because this would take their output above cap? Scheduled units are not inhibited from delivering PFR by the fact that their dispatch targets are essentially firm, so it would be odd – and unhelpful – if VRE facilities were behaving differently.
• Alternatively, if VRE facilities aren’t yet able to meet PFR requirements – why not? VRE units with surplus availability – a generally material share of the SDC on group – represent a sizeable source of energy able to resist downward frequency drift when other units under-deliver. It would be shame to waste that capability.

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

About our Guest Author