PG&E is laying the groundwork for flexible data center interconnection
Exclusive: The utility has been piloting a way of getting new load online faster — but it’s proving tricky.
Photo credit: Henry C Jorgenson / Shutterstock
Photo credit: Henry C Jorgenson / Shutterstock
PG&E has been quietly piloting flexible service agreements to get new load online faster in capacity-constrained parts of California.
Dubbed “Flex Connect,” the program allows customers that can rein in their demand during specific hours during the summer — including by using their own behind-the-meter generation — to connect to the grid in places where their full requested capacity would have been denied.
The utility currently has three Flex Connect pilot sites online, Latitude Media has learned, including a public fast charger site and front-of-the-meter batteries, the latest of which came online just last week.
According to the utility’s head of grid edge innovation Alex Portilla, the goal is to bring an additional site or two online by the end of the year, and ultimately to standardize and accelerate the process.
PG&E’s pilot comes as utilities nationwide grapple with multiplying interconnection requests and long queue times, particularly in the context of new data centers to support the AI boom. As a result, discussion has emerged of whether that new load can be flexible — though options like batchable workloads and load shifting between data centers in different locations have been at the forefront.
“This is an opportunity for us to essentially offer capacity as it is available in a granular sense, based on the actual conditions and the actual amount of grid availability,” Portilla told Latitude Media. “The alternative is a planning scenario where we essentially do a study and we say this is how much we can serve you 24/7/365 — basically limiting them to the worst-case scenario.”
Portilla said PG&E has Flex Connect sites in development across a diverse range of customers, including public and commercial charging sites, commercial batteries, and industrial sites. Eventually it plans to include small data centers as well — though Portilla noted that what PG&E is deploying today “is all on the distribution grid,” and “doesn’t apply to a 100-megawatt data center.”
For the types of massive data centers some hyperscalers are looking to build, he added, flexible connection is more of a transmission challenge, which requires a different set of considerations.
And while sites like fast chargers might choose to be flexible with their loads, most data centers likely don’t have that same level of flexibility — hence the demand for “five nines” of reliability, or 99.999% uptime. Instead, Portilla said, data centers participating in Flex Connect will likely opt for a combination of onsite generation and storage for the time periods when they need to pull back their consumption from the grid. They won’t need to disconnect entirely though; it’s more likely they would need to offset part of their load during certain peak hours.
“We don’t have any of these [data center agreements] that are in development in detail yet,” Portilla clarified. “We do have conversations happening.”
Inside the pilot
Despite the pipeline of willing Flex Connect customers, Portilla said the pilot has run into some challenges. As a result, getting it off the ground has taken longer than expected.
For Flex Connect sites to come online, they need to be able to receive and act upon automated protocol messages from PG&E that contain day ahead hourly limits. As Portilla put it, “they need to be able to receive that signal and then translate that signal into load reduction.”
Program customers must cover the costs associated with installing the load control technology that allows each site to communicate with PG&E’s distributed energy resource management system, or DERMS. Given the “significant consequences” if that communication fails, PG&E has spent months testing various scenarios, Portilla said.
For instance, if a site loses communication with the utility, it’s supposed to default back to an originally-planned consumption limit. And if there’s a power outage, a site is expected to re-establish communications with PG&E’s system as soon as it comes back online.
In part because of those testing requirements, PG&E scaled back its deployment plans for 2024; Portilla and his team initially hoped to get ten sites online this year, rather than the four to five in the works.
The three sites that are already online have been through the utility’s commissioning process for new technologies, but given that the team is still in the early days of rolling out the program, the technology didn’t get the greenlight right away. As Portilla explained, “it didn’t work just perfectly” when tested in the field for the first time, so the vendors needed to make modifications — a process that he noted isn’t unusual for the new technologies that PG&E introduces.
Now that a handful of sites have passed the commissioning process, though, Portilla expects future projects to be “much smoother.”
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Another bottleneck that has slowed down deployments is on the customer side. For example, customers applying for interconnection — whether data centers or charging sites — may be expecting relatively long lead times, and would have a decision to make about whether they want to come online in 2024 or 2025 with limited capacity. Many haven’t acquired the site yet, and are simply applying for service to see if capacity is available. In those cases, it could take another year to acquire the site, get the permitting, and build the fast chargers, for example.
“Some of the projects we planned to do this year got pushed out to next year because the customer isn’t finished with their construction jobs,” Portilla explained. “Other times the customer has decided they’ll wait to go out and build the site until the full capacity is available.” And in other scenarios, projects simply turn out not to be the right fit for the program.
Looking ahead
The current state of the program is “opportunistic,” Portilla said, meaning that when a customer applies for service and can’t secure enough capacity, his team is informed on a case-by-case basis and then reaches out to pitch the pilot. Moving forward, the goal is to be more standardized, so that when requests for service are limited, those customers are always informed about Flex Connect and their option to enroll in the program.
“We’re trying to fill the front of our funnel, essentially,” Portilla explained.
Another key goal is to determine how PG&E’s work on the distribution level can be relevant for others in the industry — via, for example, the EPRI-led coalition that PG&E joined in October focused on demonstrating flexibility.
PG&E is also conducting a cluster study of nearly two dozen data center interconnection applications the utility has received in the last year, which account for 3.5 gigawatts of additional load.
Ultimately, though, flexible service connection is just a “beachhead use case,” Portilla said. “Down the line, [this is] potentially how we might want to operate the system for more, different types of customers,” he explained, “or we may want to create new programs that allow us to effectuate changes in the load curve, either through pricing or through signals.”
The program is part of PG&E’s broader effort to maximize utilization of the existing grid, which, in addition to benefiting specific customers looking to come online quickly, will ultimately put downward pressure on overall rates, he added.
