Just before Christmas, the U.S. Department of Energy announced a new round of funding for projects that advance solar forecasting technologies. Eight projects were awarded a cumulative total of $12 million. Add the cost-share requirements for awardees, and the total value of the projects comes to $14.6 million of public- and private-sector investment.
“These projects will address a critical gap in our research, which is knowing precisely how much solar electricity to expect at any given hour on any given day,” Energy Secretary Rick Perry said in a statement.
The new tranche of funding builds upon projects funded by the Energy Department to improve solar forecasting research in 2012. One of those projects, IBM’s Watt-Sun program, used machine learning to improve solar forecasting accuracy by as much as 30 percent.
In an interview, Charlie Gay, director of DOE’s Solar Energy Technologies Office, said that solar forecasting technologies are an increasingly valuable tool in optimizing the operation of the grid as solar PV increases market share.
“The greater the accuracy, the better the grid can operate, and the more efficiently we can make use of the existing infrastructure on the grid as solar becomes an increasingly large fraction of generation,” he said.
Projects address two blocks of time, three topics
The eight projects included in the Solar Forecasting 2 funding round aim to improve grid operators’ ability to predict solar irradiance and power production. According to Gay, the projects were specifically selected to improve forecasts for two blocks of time: 24 to 48 hours in advance for day-ahead planning, when grid operators must manage projections for how to balance generation assets, especially on the transmission grid; and one to six hours in advance, when local weather events, cloud movements, or changes in cloud density can affect how much light is present.
“The intensity attributes of the light need to be well understood to connect the dots about how much light is reaching the panels and how much electricity is coming from the system,” said Gay.
The selected projects address three topics: one project will build a testing framework giving industry and academia the ability to evaluate and compare the performance of solar irradiance and solar power forecasting models; four projects seek to improve solar irradiance predictions; and three projects will examine how to integrate solar forecasting technologies with grid operators’ energy management systems.
A team at the University of Arizona will tackle the first topic. One can think of it, according to Gay, as “setting the cornerstone for quality of information,” developing a transparent set of rules, criteria and metrics for solar forecasting.
Projects bundled in the second topic aim to improve on the ability of existing tools to predict how much light, or what intensity of light, will reach solar panels. It’s all about “giving the grid operator a better handle on what to expect for the output of the plant as clouds move across the landscape,” said Gay.
The projects addressed by the third topic, he said, “put all these pieces together with the energy management systems grid operators use day in and day out.”
California, Hawaii show how to integrate solar PV
States with a high penetration of solar PV like California and Hawaii are the proving grounds for these forecasting technologies. Utility-scale and distributed solar together account for nearly 15 percent of California’s net electricity generation.
“Both California and Hawaii are leaders in putting all the pieces together,” said Gay. They are “connecting the dots between information about the grid’s performance and the projections and probabilistic outcomes that can be modeled with a set of tools that would derive from programs like the Solar Forecasting 2 program.”
He cited a collaboration between First Solar, the National Renewable Energy Laboratory, and the California Independent System Operator which proved that inverters can be used to support the operation of the grid at a 300-megawatt solar PV power plant in California, which GTM reported on last year.
“With that came modeling of clouds passing over that PV power plant, demonstrating that you could manage within a control band the voltage and frequency properties that were the standard sought by CAISO,” said Gay.
Gay stressed the real-world applicability of the projects funded under Solar Forecasting 2. The intent is for projects to be immediately relevant to the ultimate users — grid operators at the transmission and distribution level.
“These aren’t things that are done just in a classroom environment,” he said. “These are actual tools that are worked on collaboratively all the way across the value chain, including with the end-user utility grid operator.”
This article was originally featured on greentechmedia.com.