When the coronavirus power load demand destruction began appearing in Enverus Trading & Risk’s daily ISO load forecasts, our team had to quickly tackle a new set of realities.
Just like businesses all over the world, we were faced with never-before-seen scenarios as a result of the nationwide shutdowns. Our machine learning-based forecasting tools have never learned the demand dynamics at play—nor had our team. We put our heads together and quickly came up with a plan—and a way to measure the demand destruction for our customers.
First, a primer on our typical standard operating procedure: our machine learning models look at actual temperature and power demand data from previous days and weeks, learning as it goes. The models were built to respond to load growth or anything that may be happening in the market.
This presented a minor challenge. The model was suddenly forced to predict demand in a scenario where it had no historical data to compute. Our customers were understanding that it would take several days for the model to absorb new data and relearn the market. We had customers reach out and ask us how they can track the demand destruction underway.
The solution
We launched a new model that knows what the temperatures were for the load from February 2019 to February 2020. But it doesn’t know anything that’s been happening over the last month.
Below is the first look at the results of our study—just a few days since launch. Now that we are running those two load forecasts in parallel (one that’s learning the demand destruction, and one that isn’t) you will be able to see the difference between the two loads, and you can identify the level of demand destruction.
What’s happening in ERCOT, the Texas power market?
We have measured little to no demand destruction due to the coronavirus and governments implementing policies to reduce the virus in ERCOT. However, we are starting to detect some demand destruction over the last couple of days. When comparing similar pre-COVID-19 temperatures, we are detecting 1-3% demand destruction.
In NYISO, the New York power market, it’s a different story …
In NYISO, it’s a different story. We have detected demand destruction starting early last week. When comparing similar pre-COVID-19 temperatures, we are detecting 10-15% demand destruction.
Helping power market participants in a time of need
Power trading liquidity has fallen amid the uncertainty around the coronavirus pandemic. We’d like to offer you free access to this analysis in our twice-daily reports during this tough time.
We are all in this together, and we look forward to continuing to help power traders and analysts make the best decisions for their businesses, even in trying and uncertain times. Sign up here to get access to our new demand destruction load forecasts, which we are offering to you on a complimentary basis.
Latest posts by Rob Allerman (see all)
- Coronavirus Power Demand Destruction Forecasting 101 - March 27, 2020
- Deep Learning Brings Texas-Sized Accuracy to Power Forecast - March 3, 2020
- How Utilities Can Mitigate Price Risk With Enverus Forecasts - February 11, 2020
