Renewable energy weather forecasting requires more sophisticated algorithms and analysis as solar and wind penetration increases and grid operators in the U.S. are preparing for that probability.
While searching for a method to fully track the expected amount of solar and wind power outputs for real-time load balancing, technological and logistical improvements are still needed as more intermittent resources connect to the grid or are added behind the meter (BTM).
Tracking weather impacts on the ramp ups and ramp downs from renewable resources is crucial to resilience. Extreme hot or cold can lead to low wind output. And the sun isn't always shining, as evidenced by a recent island-wide blackout in Hawaii where cloud cover prevented solar recovery when a generator tripped offline.
"As you get more solar or wind, the ramps are going to be greater."
Pete Brandien
System operations vice president, ISO-New England
Solar and wind deployment is expected to continue as prices decrease and states increase their renewable resource targets. More intermittent resources also means the amount of energy ramping up and down will increase. Some grid operators, like the California Independent System Operator (CAISO), are dealing with this.
Based on volatile renewable energy outputs, the federal government dedicates resources into weather forecasting research, according to Manajit Sengupta, chief scientist at the National Renewable Energy Laboratory (NREL).
Utility Dive spoke with the people behind system operations at ISOs and Regional Transmission Operators (RTOs) in the U.S. to understand the sophistication needed to track when the sun is shining or when the wind is blowing.
Most forecasting is done with third party vendors, as ISOs and RTOs look over each other's shoulders to understand best practices when more solar or wind is deployed in a region.
"The value that we get from using a third party service is not just [their satellite based information]. Their whole core competency is using this forecasting," Joe Mulhern, senior engineer for PJM Interconnection's generation division, told Utility Dive.
Forecasting is only part of the challenge, albeit an important one. The need for generation reserves and additional flexibility, through demand response, energy storage or other technologies, is needed to ensure reliability during extreme weather events, such as the energy emergency alerts sent in Texas last week.
"A forecast won't improve the weather," Sengupta said.
Who's doing what?
California ISO
Weather Forecasting Vendor(s): Several, undisclosed, although the ISO began managing its in-house forecasting for short-term (5-minute) increments as solar capacity began to exceed 1 GW.
Upcoming: Improving visibility, especially in BTM resources. ISO is in the solicitation process for vendor services to forecast off-grid solar resources.
Electric Reliability Council of Texas
Weather Forecasting Vendor(s): AWS Truepower, Energy & Meteo Systems for wind, Maxar for solar forecasting
Upcoming: Trying to add a five-minute forecast for solar into dispatch algorithm, similar to its short-term wind forecast. DOE's short-term solar forecasting is "specifically being developed for us, but would be socialized" for the use of other vendors and grid operators, Dan Woodfin, senior director of ERCOT's system operations, told Utility Dive.
ISO New England
Weather Forecasting Vendor(s): Several, undisclosed vendors supply ISO with data for its load forecasting models; DNV GL provides additional wind and PV output forecasts, contributing to the net load forecast.
Upcoming: Anticipating the need for forecasting offshore wind output, evaluating the need to integrate BTM resources.
Midcontinent ISO
Weather Forecasting Vendor(s): Third-party vendor develops short-term wind forecast, while MISO uses load forecasting software (that combines recent system load and expected weather conditions) for overall short-term load forecasting.
Upcoming: Currently investigating the value of having multiple vendors and options to gain visibility on BTM resources; looking at what other grid operators are doing with solar forecasting.
New York ISO
Weather Forecasting Vendor(s): AWS Truepower for wind and solar forecasting; additional partnership with New York State Mesonet, focusing on connected and BTM solar.
Upcoming: Continuing to improve capability to collect information, and improving existing tools with the state utility regulators and energy agency.
PJM Interconnection
Weather Forecasting Vendor(s): An undisclosed wind power forecast vendor in use for more than a decade, and an undisclosed solar power forecast vendor in place for about three years.
Upcoming: Integrating the data from BTM resources, exploring real-time BTM data to validate existing models; anticipating the need for forecasting offshore wind output.
Southwest Power Pool
Weather Forecasting Vendor(s): Several undisclosed vendors for weather and renewable energy forecasting, focused on their growing amount of wind resources.
Upcoming: Continuing to update processes, criteria and procedures, based on the large amount of wind penetration in the system — nearly 23% at the start of 2019.
Tracking cloud cover
Weather forecasting has evolved from alerting RTOs and ISOs about large storms and natural phenomena to tracking the amount of UV radiation that will be available in real time to power up solar systems.
However, national weather service statistics aren't going to provide the amount of solar on the grid in real time, according to NREL's Sengupta.
The "state of the art forecast" for solar that vendors provide is typically adapted from the National Oceanic and Atmospheric Administration (NOAA) forecast. "NOAA's goal for a solar forecast is not solar radiation," Sengupta said, making vendor expertise and resources more important.
DOE invests in capabilities for tracking solar radiation, which includes efforts from NOAA on DOE-funded projects meant to "essentially improve the models almost to the point where they can do a better job of predicting solar radiation for these big plans," Sengupta said.
As "clouds at different fronts move across the state, it's going to impact different solar installations at different points in time," Emilie Nelson, New York ISO's executive vice president, told Utility Dive.
"It's important to capture some of that variability at different points in time" and on a granular level, she added, in order "for the system to manage around that variability.
"We need an accurate idea of what the conditions are [for intermittent resources on the system]. One of the inputs are where the load is going to be and where renewables are going to be," Nelson said of NYISO's five-minute real-time load balance forecast.
Grid operators use weather forecasting to communicate to members how much other generation needs to be dispatched in real time to guarantee reliability.
"There's a lot of things that we're learning from California," which has a lot more renewables on their system, Peter Brandien, ISO-NE's vice president of system operations, told Utility Dive. "It's not just necessarily about forecasting the output from these resources and ensuring that we have the right resources for the load ramps that we would see."
Geographic diversity "reduces overall forecast error from a percentage basis," ERCOT's Woodfin said.
"What we saw with the build-out of the wind generation is that ... the first 1000 MW or so was in a fairly tight geographic area," Woodfin said. Therefore, if a cold front came through, decreasing the amount of wind, "would affect basically 100% of the [wind] generation."
This added a lot of pressure for ERCOT's day-ahead forecast if the ramping or lower output from the cold front was delayed by an hour or so from the original forecast, he said. "Once you start to get the wind more geographically dispersed for a broad area, if some weather pattern happens a little slower ... they balance themselves.
"When you consider intermittent resources such as wind and solar there is a general concept that having a broad geographic area provides a diversity of the output of either wind or solar," NYISO's Nelson said. "When you think of a single state the size of New York, some of the diversity that we see as possible becomes more a function if you have land-based or onshore resources versus offshore."
New York's wind farms are mostly in the northern and western parts of the state, while NYSERDA awards offshore wind contracts for leasing areas off of Long Island.
ISO-New England is also facing offshore wind build-out in a small geographic area. "Generally the weather fronts that come across New England touch all of New England," Brandien said.
Market size determines sophistication
As solar penetration grows regionally, grid operators have more volatility to account for, several experts said.
"As you get more solar or wind, the ramps are going to be greater," Pete Brandien, ISO-NE's vice president of system operations, told Utility Dive.
ISO-NE's six state region is relatively small geographically, but the 3,000 MW of solar deployed in the region is expected to increase based on state incentives, Brandien added.
"There could be vendors who are in many ways processing the data" for solar forecasting from satellites, Sengupta said. "The tipping point in terms of investment" becomes the growing need for responsive operations as the amount of resources grows.
"Ultimately, to get [high resolution] satellite data, the government puts a multibillion dollar effort into these satellites. People like IBM or other big companies can provide these data without much latency" between when RTOs/ISOs give them information and they return a forecast, Sengupta said.
Enter CAISO, which exceeded 10 GW of grid-connected solar in 2016.
"One of the challenges that we realized is that regardless of how good [or] accurate your forecast is, the fact that we have to deal with processing of the data, ... created what we call a latency problem," Guillermo Bautista Alderete, CAISO's market analysis and forecasting director, told Utility Dive. "And that is regardless of how accurate your forecast is. It's just a matter of timing the unit from end to end to bring the forecast into our system."
Over the last two years, CAISO created in-house capabilities for a five-minute real-time solar forecast to "shortcut all processing time and reduce the latency issue," Alderete said. That became operational earlier this year. "We have a mix of external vendors to provide forecasts in the longer horizon, but for the very short term, we rely on an in-house forecast that we're developing internally."
However, some grid operators have sought big changes to their weather forecasting without having as much solar as CAISO. NYISO has about 1500 MW of solar in the region and partnered with Mesonet and the University of Albany to increase sensors for UV radiation on a more granular level.
"The Mesonet gives you a somewhat local immediate view," Sengupta said. The program was launched in 2014 by Gov. Andrew Cuomo.
"It is important now to have this information and it's going to become more and more important as the volume of solar resources increases into the future," Nelson said. Those tools and expanded capabilities will help the grid operator "respond to the reliability need."
Behind the meter and beyond
Some grid operators, including Midcontinent ISO (MISO), are at their infancy in determining what to do with distributed resources.
MISO works to understand state-level policies, as its members are in fully vertically regulated states, J.T. Smith, MISO's director of economic and policy planning, told Utility Dive. MISO looks to grid operators with higher penetrations of solar, especially BTM resources.
In CAISO, "all that we see is the final effect on the load" from BTM resources, Alderete said. The grid operator began a competitive solicitation process this year for BTM solar forecasting due to the "increasing need for visibility behind the meter."
PJM has gained visibility into BTM resources through its subsidiary, the Generation Attribute Tracking System, an online database where owners register their solar systems to qualify for renewable energy credits within their state.
"With solar, we found that there's a growing number of BTM solar farms," David Schweizer, PJM's generation manager, told Utility Dive. "The challenge has been gaining the visibility and awareness of those BTM solar farms in order to integrate them in our solar forecast."
Individuals and non-participating members send real-time data to PJM on their BTM systems. Other limited public data is available, and the company wants to increase real-time visibility of the systems to ensure the viability of their load-balancing forecasts.
The grid operator is working to better understand the impact of BTM solar on the market as individual and larger unconnected systems continue to be developed, while "increasing operational awareness but also potentially integrating" BTM resources into PJM's outage dispatch, Schweizer said.
CORRECTION: A previous version of this article incorrectly referred to the hot weather events in Texas during August, 2019. The grid operator had announced unplanned outages and issued two energy emergency alerts (EEA1).
CLARIFICATION: DNV GL uses weather forecasting to provide ISO-NE with wind and solar output, but third-party vendors are used for the grid operator's weather forecasting.