Last week brought yet another blast of cold temperatures – the third this year! With March around the corner, we can’t help but wonder: Was this winter’s last cold snap? For most of the U.S., March and April signal the arrival of spring, and many are already feeling the warmer temperatures this week, which typically means weaker energy demand, generation outages, and increased curtailments.
But before we move on, let’s take a look at how Amperon’s forecasts performed during what we hope was the final polar vortex event of the season. The deep freeze from February 14-20 pushed energy demand to new winter peak records in ERCOT and SPP, with prices spiking above $1,000/MWh in some regions. In ERCOT, temperatures on Wednesday were colder than expected, leading to a demand miss almost entirely driven by the unexpected drop in temperatures across the state.
ERCOT
In ERCOT, temperatures were as much as 5-6 degrees colder than forecasted on Wednesday, February 19th, leading to a demand miss almost entirely driven by the unexpected cooler temperatures across the state. EROCT’s day-ahead forecast saw this major demand miss with a 14.93% APE at HE12 on Wednesday the 19th. But when looking at forecasts with the actual weather, Amperon’s Weather Actual Demand forecasts were 78.4% more accurate than ERCOT.
Thursday the 20th set a new winter peak demand record of 80.4 GW, breaking the previous record of 78.1 GW set on January 16, 2024. Amperon’s accuracy for that hour was 0.46% at HE8. That’s 84.5% more accurate than ERCOT’s APE of 2.97%. Six days out, Amperon started forecasting demand with forecasts closely matching actual demand and accuracy as low as 0.93% APEs at the peak demand hour. The final MAPE for the day with the highest demand peak on the 20th was 1.39%, which was 45.6% more accurate than ERCOT.
A quick look at our Day Ahead Price forecasts for February 20th showed good results battling the cold snap. The magnitude of HE8 (peak demand hour) was reached. HB_Houston HE8 forecast was within about $20 of what DA prices ended up clearing at. HB_North and HB_South forecast were within about $100 of what the day-ahead prices ended up clearing at.
Here’s a quick roundup from the other regions:
- PJM: The 19th was PJM’s coldest day of the week seeing demand peak around 128.4 GW – although this cold snap was not as extreme as what occurred at the end of January for PJM. At the peak, Amperon’s APE was 0.75%, 58.55% more accurate than PJM’s APE of 1.81% at this time. For the whole week, Amperon was 19% more accurate than PJM.
- MISO: The 20th saw the highest demand of the week just shy of 105 GW. Similar to PJM, last week’s cold snap was not as extreme as what occurred at the end of January for MISO. We still outperformed the ISO with 46% more accurate forecasts during the coldest days of February 18th-20th. For the whole week, Amperon was 55% more accurate than MISO.
- SPP: Thursday, the 20th set a new winter demand peak record of 47.9 GW for SPP, which broke the previous record in December of 2022. On this day, Amperon’s MAPE was 0.82% – 62.2% more accurate than SPP’s MAPE. For the coldest days, (17th-21st) Amperon was 42.7% more accurate than SPP.
- ISONE: Tuesday, the 18th at HE19 saw the highest demand peak of the week. Amperon APE was 0.40%, which was 91.34% more accurate than ISONE APE of 4.62% at that hour.
- NYISO: Tuesday, the 18th saw the highest demand at HE19, and our APE was 1.02% – 49.75% more accurate than NYISO’s APE of 2.03%. Using Amperon’s weather actuals data, rerunning the model with actual weather to eliminate the weather miss at the hour resulted in an even lower APE of 0.63%.
After a cold week, most of the country is experiencing spring-like or warmer-than-normal temperatures this week. ERCOT, PJM, MISO, SPP, NYISO, and ISONE are currently seeing their warmest days of the week, with some regions seeing a mix of winter temperatures returning into the weekend. But whether it’s a cold snap with record-breaking demand events or temperate weather with weak demand levels, having accurate energy forecasts is critical to your operations
.png)
.png)
