The Bonneville Power Administration may have an enviable grid system, with the country’s largest hydropower resources providing the Pacific Northwest region with cheap and plentiful electricity.
But that doesn’t mean the federal power agency isn’t facing challenges.
Saving the Columbia River system’s salmon stocks has placed severe constraints on how dams can be tapped as flexible power resources. Meanwhile, Bonneville is balancing about 4,700 megawatts of intermittent wind power in its system, and it has a growing share of solar power to manage as well.
These changes are pushing Bonneville to explore new options for demand response, or curtailing power loads to meet its needs related to grid balancing, transmission congestion, and other issues.
Unlike the day-ahead and hour-ahead demand response common in the rest of the country, however, Bonneville wants its resources to be able to respond in 10 minutes or less. It also wants to build a platform that can pull in multiple resources, from multiple public and private aggregators, in the same control rooms where it dispatches its hydropower resources and participating power plants.
Last year, Bonneville hired Redwood City, Calif.-based startup AutoGrid Systems to design this platform. This week, it unveiled the results -- a software system that since February has dispatched some 20 events to shed between 18 megawatts and 30 megawatts at a time, for a total of 500 megawatt-hours of grid-balancing load drop.
That’s a relatively small amount, compared to the gigawatts' worth of demand response at play in the markets run by mid-Atlantic grid operator PJM, ISO New England and New York ISO. But it’s an important foundation for Bonneville as it seeks to determine how much of the roughly 1,000 megawatts of potential demand response in the region will suit its needs.
“To succeed, demand response has to meet three objectives: it has to be highly reliable, it has to be cost-effective, and it has to be easy to use and deploy,” said John Wellschlager, a Bonneville account executive working on the agency’s demand response initiative. “While DR is pretty well understood as a concept, it’s an immensely complicated product to be used as a reliability tool.”
A fast-acting, cross-platform approach to demand response
He’s talking about using demand response for ancillary services, which call on generators or other grid assets to quickly ramp up generation or ramp down consumption. In the past few years, Bonneville has launched three different demand response pilots -- one with the city of Port Angeles, Wash., another with demand response aggregator EnerNOC, and the third with Energy Northwest, the entity that owns and operates wind and hydro assets, as well as the Northwest’s only nuclear power plant, for 27 public utilities serving more than 1.5 million customers.
Results have been mixed. The first project, with the Nippon Paper mill in Port Angeles, used phone calls to mill operators to power down loads, and while “they worked very hard” on meeting Bonneville’s needs, “a single point of failure on a load like that didn’t provide us the kind of reliability we needed,” he said.
“Aggregation is the future of DR, not single-source loads,” he said. “You need to have some redundancy to create the level of reliability you’re looking for.” That conclusion informed Bonneville’s projects with EnerNOC, which is tapping loads like food-processing and cold-storage facilities, and with Energy Northwest, which has enlisted two paper mills, a small-scale battery bank, and a conservation voltage reduction system run by the municipal utility of Richland, Wash.
These projects use the digital communications protocol OpenADR to assure a much faster dispatch and response of end loads. But Bonneville wanted a single interface to be able to control both sets of demand response aggregations, as well as the ability to add new resources in the future, he said.
“We started working with AutoGrid” in November 2014, he said. “The implementation went about as smoothly as we could have hoped. We’ve had 24 test events, and not one failure. And we’re hitting them at all hours of the day, within shift changes, in the middle of the night, and they have to deploy within 10 minutes.”
Cara Ford, Bonneville senior project manager, said the key requirements for this platform were speed and flexibility in integrating different demand response portfolios with the agency’s operations center.
Bonneville’s grid operators “are managing the flow of the river and the reliability of the power system,” she said. “They needed to have one unified interface to dispatch multiple programs. While we could just use Energy Northwest’s system, or just use EnerNOC’s system, we’d have two different interfaces to learn. We needed somewhere to grow the system and dispatch multiple products.”
Bonneville also needed software that could be hosted on a cloud computing platform. “We’re in a demonstration phase, where all the requirements aren’t known,” she said. “This is a way for us to start experimenting with using demand response, and having a platform without making that investment” in hosting its own software.
AutoGrid’s Demand Response Optimization and Management System (DROMS) is playing a role in projects at utilities including Florida Power & Light, Hawaiian Electric, Oklahoma Gas & Electric, and Austin Energy. The startup has received funding from ARPA-E, venture investors and German utility E.ON, and specializes in ingesting, organizing and analyzing disparate sources of data at speeds required for time-sensitive grid operations.
In the case of its Bonneville platform, AutoGrid is projecting how much load is available to be shed, sending the OpenADR signals to the participating entities, and then verifying the load reduction, on “almost a minute-by-minute basis,” said Raj Pai, the startup’s global head of products and marketing.
While this task is relatively simple with a handful of loads that face strict penalties if they don’t respond to Bonneville’s dispatches, AutoGrid’s software is built to expand to a much larger set of endpoints, and target dispatches to certain locations to help reduce transmission congestion on specific parts of Bonneville’s grid, he said. It’s also designed to churn through historical and real-time data to get a finer fix on just how much load shed can be expected from less firmly controllable endpoints, such as lots of homes equipped with smart thermostats.
That’s important for Bonneville as it moves on to the next phase of its demand response plans, Wellschlager said. “At this point, we want to explore what buckets of DR are going to best meet our needs,” whether it’s for augmenting power plants that supply balancing capacity, reducing temporary transmission constraints during heat waves, meeting winter peak load events, or even increasing consumption when wind and hydro power is generating more than the system needs.
The Northwest Power Council, a four-state entity that creates regional power plans, has identified some 1,000 megawatts of fast-responding demand response potential in Bonneville’s region, he said. But “the question is not how much DR is in the region, but how much is cost-effective to use as an alternative to generation.”
As it continues to expand its pool of available demand response resources, “we’re going to use the platform with EnerNOC as well as with Energy Northwest,” he said and “we’ll be able to use the same platform, no matter what aggregator is bringing load to us.”