P30 demand forecast glitch hunches

It’s hart to resist making a hunch on the reason(s?) for the Glitch in P30 demand forecast for QLD on Tuesday evening 29th October 2024.

To sum up the glitch:

  • There was one quirky forecast issued at 18:30, for the 19:00 half-hour period and beyond.
  • Following this, the next three forecasts were all “flat”, until about 04:30
    • Flat until 04:30, first 30-minute period of the next trading day, where they meet up with the earlier forecasts.

Here’s a complimentary view, using the ‘chart’ mode of ez2view’s Forecast Convergence (the vertical dotted line represents the point in time of 20:45):

We received a suggestion to look into the PV forecasts (rooftop , we assumed).

Yet for the rooftop PV forecasts to be in error by the equivalent amount we’d need cover the MW difference charted above – more than 2,000 MW. But we don’t observe that in the chart of the rooftop PV forecasts, below.

So we conclude, on this occasion, the rooftop PV forecasts don’t explain the demand forecast glitches. Moreover the glitches were from around 7pm onwards – after daylight hours in QLD.

 

With PV off the table for now, I thought I’d offer my own hunch: That the temperature sensitive component of demand in the model somehow went missing.

Purely by visual inspection, the three main glitchy forecasts all appear to take a generally flat trajectory. They sit at a level equivalent to the overnight minimum at around 04:00.

If we took simplistic view of the demand forecast as including a component tied to weather-dependent human activity, and a component tied to ‘usually always on’ demand.

The latter component may still change with time but not respond to weather.

The human component would be temperature-sensitive, and generally fall away to a small level when most people are asleep – hence the overnight minimum that we typically see would be driven by the ‘usually always on’ demand.

If the human component was missing from the forecast, might we get a flat forecast sitting at that low 4am level? Could this be the reason?


About the Author

Linton Corbet
Linton joined Global-Roam as a software engineer and market analyst in August 2020. Prior to joining us, he worked with the AEMO for 7 years, and before that, as an air quality scientist.

Be the first to comment on "P30 demand forecast glitch hunches"

Leave a comment

Your email address will not be published.


*