By Elizabeth Anastasio, senior meteorologist – Generation
Power plants and transmission lines usually grab the headlines in energy industry news, but another component is equally integral to PJM’s mission of ensuring that power flows to our customers wherever and whenever they need it.
That element is forecasting customer demand, also referred to as “load” – which means, basically, predicting human behavior. PJM does that by analyzing a number of factors that influence electricity use, including the weather, season, holidays, day of the week, time of day and special events like the Super Bowl. We also look closely at historical data to see whether previous patterns are likely to repeat themselves.
Accurate load forecasting is key to procuring and scheduling enough power to meet customers’ needs – but not so much power that unnecessary expense is incurred. The effect of load forecasting ripples through all of PJM’s operations and planning decisions, from how much generation should run today, to what system investments are needed into the future.
How Weather Impacts Load Forecasting
Weather has a big impact on load forecasting, because it dictates customer behavior when it comes to heating and cooling needs. With the growth of wind and solar generators, weather also has a significant effect on the amount of energy generated by wind and sun. We have to look not only at temperature, but also characteristics such as wind speed and cloud cover.
For this reason, PJM employs its own staff meteorologist. That’s me.
Remember the solar eclipse from Aug. 21, 2017? That’s a good example of how all of these factors can combine to affect customer demand. At first, we were predicting a higher load, knowing that a number of solar panels would go dark along with the sun. Then, other elements of the weather – notably cloud cover and cooler temperatures – reduced cooling needs. And finally, human behavior stepped in, as people went outdoors to watch the eclipse, lowering the demand for electricity even further. This event also provides a valuable learning experience for the next full solar eclipse in April 2024.
This example also underscores the importance of having meteorological information updated as frequently as possible – something that’s improving along with advances in our nation’s weather satellites. PJM dispatch uses publicly available data sources, such as the National Oceanic and Atmospheric Administration, in combination with PJM’s personalized suite of data tools, to continually refine the load forecast. They predict energy use, throughout the day and six days forward, throughout nearly two-dozen distinct zones within PJM’s 13-state footprint.
While we employ plenty of computer-assisted modeling, much of our work is manual, and one of our most valuable tools is the experience of our dispatchers. This is part of what makes the PJM load forecast unique.
Each model has strengths and weaknesses, and you can run them and come up with different solutions. At the end of the day, it’s up to the dispatcher to choose the best load forecast.
Incorporating ‘Behind-the-Meter’ Generation into Forecasts
By definition, forecasting looks to the future. But its accuracy actually relies on the past. Forecasts are always better when you can draw from historical performance.
This is one of the challenges of incorporating what’s called “behind-the-meter” generation – particularly solar – into the load forecast. Behind-the-meter generation, as the term implies, is power that’s not being generated on the grid or measured in a way that we can see as the grid operator.
Over the past 10 years, the biggest increase in solar generation has been in behind-the-meter installations, like common rooftop solar panels. It’s critical to include these resources in the forecast, because while the generation isn’t connected to the grid, the customers it serves are. So when the resources stop generating, the gap in power can lead to steep ramp-up times for the grid-generated power that must replace them.
There is a lot we don’t know about these installations, starting with historical performance. We are told only the output capability and location of solar panels when they are activated. But PJM isn’t able to track, for example, whether they’ve become less effective over time from wear or grime, for example.
It’s also difficult to predict solar generation output, because it is dependent on cloud cover – an aspect of weather that can be localized and harder to predict far into the future.
Increased penetration of behind-the-meter generation highlights our need for more detailed, granular data. We can no longer assume that the load demand on a single power station closely follows that of the whole PJM footprint, or even the zone that it’s in.
Since initiating behind-the-meter load forecasts in 2016, PJM has made progress in integrating that data into our load forecasts. We have made it a priority to further explore the impact of behind-the-meter generation on the grid. As technology and the power industry evolve, so does PJM. We always seek to improve our load forecasting to help us deliver affordable power, reliably, rain or shine.
