In 2019, solar power accounted for nearly 15 percent of the renewable energy generated in the U.S., according to the U.S. Energy Information Administration. The Center for Climate and Energy Solutions expects that figure to grow to 48 percent by 2050. This trend is driven in part by states’ growing commitment to fighting climate change by transitioning from fossil-fueled power to renewable forms of energy.
As a result, independent system operators (ISO) of the national power grid are becoming increasingly reliant on solar forecasting tools to help maintain a reliable, balanced and economically efficient grid.
“Having an accurate sense of what solar panels are going to be producing helps us ensure that we have enough electricity on hand to meet our power demands,” said Matthew Kakley, a spokesman for ISO-New England. “It also helps us create the energy forecast that we use to maintain a stable wholesale market for power buyers and sellers, and ensure that we’re getting the electricity that consumers need at efficient and competitive prices.”
Solar forecasts — most of which are produced by commercial vendors — are also used by energy traders, and operators of large solar plants to manage operations or to schedule maintenance. Typical power sellers include energy producers such as power plants or solar farms. Typical buyers include local utility companies that distribute power to consumers.
Forecasting the Future
According to the National Renewable Energy Laboratory (NREL), in Golden, Colorado, there are two basic approaches to solar forecasting:
- Physical methods, which use numerical weather prediction (NWP) models to forecast solar irradiance — solar energy per unit area falling on a horizontal surface — in a particular region. Irradiance data is used to predict the power output from solar photovoltaic cells; and …
- Statistical methods, which use historic and real-time generation data to correct results derived from NWP models, or to produce short-term forecasts.
Gathering the Data
Irradiance data, explained Bri-Mathias Hodge, chief scientist for the NREL’s Power Systems Engineering Center, is typically collected either terrestrially using an instrument called a pyranometer, or from space by satellites such as NASA’s family of GOES-R environmental sensing satellites.
“Irradiance has two components: direct normal irradiance such as sunlight shining directly on a solar cell, and diffuse irradiance, which is sunlight that’s been reflected off clouds and particles in the atmosphere,” Hodge said.
Among other things, Hodge added, the geostationary GOES-R satellites can identify the types of clouds present, how they’re distributed in the sky, how much shadow they’re creating over solar farms and where they’ll move next. The satellites’ focus on short-term phenomena makes this data most valuable for predicting solar energy production within, say, the next four hours.
Adding Magic to Experience
According to Hodge, most solar forecasts used in the U.S., are based on weather prediction models — with a little bit of experience and “secret sauce” added in by forecast vendors.
“Grid operators typically use what we call a deterministic or point forecasts to predict solar output at a given location at a given time, for example, 50 Megawatts of power at 1 p.m.,” Hodge said. “If I’m forecasting a day ahead, I would use NWP model output like NOAA produces, then adjust that data based on local weather information and the history of how accurate it’s been. For an hour-ahead forecast, I’d rely more on a statistical or machine-learning approach.”
And, like weather reports, Hodge added, the accuracy of solar forecasts tends to follow an exponential decay curve.
“Once you go past four hours, the most reliable models will be the NWP variety,” he said. “If you’re looking at the weekend or a week ahead, you would still use an NWP model, but the accuracy of a week ahead is much less than it would be for say, a day-ahead forecast.”
Improving the Odds
To make solar forecasting more reliable for national power grid operators, Hodge advises, the solar industry is now moving toward using more probabilistic forecasts.
“There will always be weather-related uncertainties associated with even the most robust solar power point forecasts,” he said. “A probabilistic forecast is better at quantifying these uncertainties and the degree of risk associated with a given forecast, in the same way that a weather forecast might say there is a 10 percent chance of rain today.”
Measuring Invisible Power
In some parts of the country, such as New England, much of the solar power being consumed is not readily visible to or controlled by grid operators. This “behind the meter” (BTM) solar power comes from small, privately owned, often rooftop solar installations and is typically consumed where it is generated. At last count, advised Kakley, there was more than 3400 MW of BTM solar capacity installed in New England or about 15 percent of the region’s typical peak electrical load of 23,000 MW.
“Knowing when that solar power is going to be available and when it’s not is vitally important for our forecasting team because on a cloudy day, we need to make sure we can replace that 3,000 or so MW of capacity,” he said. “This task is especially challenging because residential solar shows up on our grid not as a source of power, but rather as a lack of demand.”
Nonetheless, added Kakley, solar forecasting remains a vital part of ISO-NE’s management of two types of energy trading markets: its day-ahead market and the real-time market. The day-ahead market allows energy traders to secure prices the day before delivery while avoiding market price volatility. The real-time market allows market participants to buy and sell wholesale electricity to balance the difference between day-ahead energy commitments and real-time demand for and production of electricity.
“We do our solar forecast in five-minute increments and feed it into the overall energy forecast that guides system operations,” he said. “Knowing how much solar is being used on the grid helps ISO-NE operators decide which other types of power resources to dispatch (i.e. select) to satisfy fluctuations in demand for electricity.”
Over the next 10 years, Kakley added, “ISO-NE expects the amount of BTM solar power in New England to double, so it’s going to be even more important to have an accurate accounting of how much power those solar panels are producing as part of developing our daily energy forecasts.”
Keeping the Lights On
From a broader perspective, Kakley advised, solar energy and solar forecasting tools are both important parts of a major transformation occurring in the power industry. U.S. cities and states are looking increasingly to renewable energy and new energy-saving technologies as a way to reduce their carbon footprint and their reliance on greenhouse-gas-emitting fossil fuels.
“There are a lot of moving parts in this transition of the power grid,” said Kakley. “Through it all, our number one job is to make sure that our system works, that it remains reliable and that the lights stay on.”
If you’re ready to add power to your career and some green to your future, please visit www.northropgrumman.com/careers.