08 Jul A Tale of Two Forecasts

By Trevor Su

Early summer is a good time for Ontarians. Temperature is warm but mild, the lawn still green. I enjoy riding my bicycle outdoors and it feels great to do it without wearing thermal layers.

Being in the energy analytics business, it also means prime time for producing global adjustment five coincident peak (5CP) alerts for our customers.

By looking at past weather and Ontario Demand data from the IESO over the last 8 years, it is evident that there is a strong correlation between weather/work-day vs weekend/hour of day and electricity demand. Historically, most 5CP peaks happen on hot summer days, during workdays, between 4 pm to 7 pm. Over the past week we have issued a few alerts to our customers and channel partners. Wednesday, July 3 – Friday, July 5 were humid warm days with real feel > 35°C. These three days are sitting atop of the 5CP on the IESO Peak Tracker.

The peculiar case I want to talk about here is Saturday, July 6. Last Friday July 5 at 6:39 am, our AQEW forecast model predicted Saturday would be a peak day. Based on our AI algorithm, the projected AQEW value is within the top 5 to-date. In addition, the long-term weather forecast predicts a mild summer. This means there is a higher probability of current peak days to remain peak till the end of the base period.

In the last 10 years no 5CP peak day has ever happened on a Saturday.

Having said that, during the last two base periods, three weekend days have ranked among the top 10 demand days, all very close to making the top 5. Predictions are not dependable on their own. Calling a peak day relies on knowledge to derive the probability of forecast.

Table 1 – Last two base period top 10 demand days

Note: January 6, 2018 ended up being 6th in AQEW (AQEW is the metric that determines 5CP and a Class A participant’s peak demand factor: AQEW = Ontario Demand + Embedded Generation – Loss)

On Friday at 6 am, the IESO’s Ontario demand peak forecast for Saturday was only at 17825MW. This means there is a roughly 2200MW difference between our forecast and theirs. It is not unusual for different models to diverge depending on input vectors, nevertheless the magnitude of the difference raised a lingering question with me.

I checked the IESO forecast on Friday at around 5 pm before I left work, and IESO’s peak forecast for Saturday was at 18233MW, hardly a peak qualifying number.

On Saturday morning at 6 am, we reiterated our 5CP peak alert position, while the IESO’s forecasted number of 19485MW did not make it to their 5CP list. Powerconsumer made the call based on its own forecast value and probability.

As of the time of this blog, Saturday’s actual Ontario peak demand ranks 4th on the 5CP list. We are interested to see if occasional weekend peaks will become the new normal.

For those who are watching peak days, tomorrow Tuesday July 9 is a low possibility, while Wednesday and Thursday look to be strong possibilities.

Enjoy the heat!