“Scheduled Demand” in Victoria stopped short of 8,500MW today (lower than forecast) but the demand peak forecast for NSW tomorrow has grown to almost 13,000MW

Here’s a brief post this evening to follow on from Monday evening’s look at forecasts for tomorrow in NSW, and yesterday’s quick note about forecasts for Victoria today to look at what actually happened in Victoria, and update the forecasts for NSW:

(A)  Actual level of demand in Victoria was lower than earlier forecasts predicted

In the snapshot caught from NEM-Watch yesterday afternoon, we see that this afternoon’s electricity demand was forecast to be up around 9,000MW.  In comparison, this snapshot from NEM-Watch at 17:50 shows that Scheduled Demand only peaked up at 8,455MW (as dispatch target) in the 16:40 dispatch interval (NEM time) late today in Victoria:

What’s more, the price did not rise above $150/MWh at all through the day.

Using the Forecast Convergence widget within ez2view, we can see how the forecasts trended downwards as we approached real time:

(B)  … but we see forecast for NSW tomorrow almost 1000MW higher than Monday evening’s level

In this NEM-Watch snapshot from 18:00 today we can see how tomorrow’s forecasts are now incorporated into predispatch, and that the demand forecast in NSW has increased to almost 13,000MW (12,986MW in the 16:30 trading period, NEM time).

Here’s the same chart from the Forecast Convergence widget in ez2view focused on NSW – in which we can see the continued trend upwards on the forecast for tomorrow afternoon:

Someone asked, in response to Monday evening’s post, what might happen to prices as a result.  Given that we’re now in the predispatch window we can look at how AEMO’s successive forecasts are trending (again in Forecast Convergence within ez2view):

In this series of overlayed charts, we see that there are forecasts that afternoon prices might reach the cap payout price of $300/MWh – however this was also forecast for this afternoon, and did not eventuate.  As I have noted before, forecasting is a mug’s game (a thankless task where the modeller is almost certain to be wrong – but might produce some useful insights anyway, and is very necessary).  This certainly applies to demand forecasting.