What if IT could be used to eliminate the West Coast’s notorious rolling blackouts or huge regional power outages like those experienced by the Northeast and Midwest in 2003?
Researchers at the U.S. Department of Energy’s Pacific Northwest National Laboratory (PNNL) in Richmond, Wash., decided to find out.
With IBM as a partner, they built a demonstration network called GridWise that showed how an event-driven service-oriented architecture (SOA) can be used to build a power marketplace that lets residential and commercial customers change their electricity consumption nearly in real time, based on price and other factors.
During the yearlong, Energy Department-sponsored marketplace demonstration, customers spent less money on power, and utilities easily accommodated spikes in demand without affecting service levels.
The marketplace, an SOA application ran on an IBM WebSphere Application Server at PNNL and received data in real time from various Web services about electricity’s current wholesale price and most recent closing price, as well as whether those prices were trending up or down.
It communicated with specialized, “smart” appliances at participants’ sites via IBM-developed middleware built within what IBM calls its event-driven architecture (EDA) framework and running on the WebSphere server. The EDA middleware provided the link between the transaction-oriented marketplace and the more physical world of the controls-based appliances.
“We found that through this system, we could influence people’s loads in a very rapid fashion,” says Rob Pratt, program manager for the GridWise project. “And that really helps out some of the more difficult tasks in managing fluctuations and potential outages in the power grid.” In addition, customers saved an average of 10 per cent on their electric bills.
How it worked
GridWise served Washington’s Olympic Peninsula, a region that relies primarily on electricity for power and heat. It began in January 2006 with 100 residential customers, one commercial building and one water-pumping station.
For the marketplace trial, customers received new electric meters, as well as thermostats, water heaters and other smart appliances, from Invensys Controls. The meters and appliances communicated their settings every five minutes to a central Invensys gateway at each customer location.
The gateway, in turn, sent signals over a ZigBee-based, wireless-sensor network to the appliances to adjust their settings based on customer-defined temperature preferences. Customers set those preferences — 72 degrees during the day and 65 degrees at night, for example — via a Web portal.
They also could set thresholds for raising or lowering their thermostat settings based on the cost of electricity: If the price of electricity rose, some customers were willing to pay the higher cost to keep comfortable, while others decided to adjust their thermostats to a lower setting to save on their bills.
“We put customers in control of their energy consumption,” Pratt says. “They’d get on their computer to program their thermostat and set their preferences, and they could just set it and forget it.”
But the Invensys appliances could not understand pricing and consumer preferences — information needed to submit bids to the PNNL marketplace — so the controls-based EDA middleware was critically important. Virtual thermostat and water-heater Java objects in the middleware added the intelligence about pricing and could interpret a user’s preset preferences.
“Using event-based programming, we bridged between the control-systems world and the SOA-transaction world,” says Ron Ambrosio, manager of Internet-scale control systems at IBM. “It let us build applications that are more control-like.”
Via Web services, the virtual thermostats would bid a certain price into the marketplace based on the current temperature in the house, what the user’s preferences were, and how responsive they wanted to be to changing prices.
Every five minutes, the marketplace would take those bids and determine a new clearing price for electricity. The new price would then flow out from the SOA marketplace through an event bus to all the virtual devices, kicking off their reaction.
“All devices would receive that same event, but they would react differently based on their consumers’ preferences,” Pratt says. If the closing price from the marketplace was at or below its bid, the virtual thermostat or water heater would accept the price and keep everything level. If the closing price was above its bid, the virtual thermostat or water heater would send a signal to the gateway in the home to adjust the Invensys thermostat so it would not use as much electricity.
“We built a very highly distributed application,” IBM’s Ambrosio explains. “Each of these thermostats was a Java object running in its own thread and interacting with the market through asynchronous events that were flowing through our event-driven programming framework, through our event bus, and into the SOA world, which is where the whole market mechanism sat.”
The result was that PNNL could manage demand more evenly. When demand increased, PNNL “ended up raising the spot price of electricity on a five-minute basis and acquiring enough reduction in load to manage a constraint,” says Pratt, explaining that the constraint could represent a substation trying to defer a half-million-dollar upgrade for some number of years, for example. “The substation has reached its limit, new customers are coming in, and you use this approach to limit the consumption of the customers that substation has to serve.”
Because this was just a demonstration project, PNNL didn’t actually change electricity rates, which would have required negotiations with public utility commissions (PUC). Instead, PNNL set up an Internet account for the participants and made a deposit that was equal to 20 percent of their previous year’s electric bill.
“We told them they could keep all that money if they played the game with us,” Pratt says. “We billed that account against the real-time floating prices, and they got to keep whatever was left in the account at the end.”
The results were powerful. “In one case, we were able to reduce consumption by 50 percent for four straight days,” Pratt says. “Plus, we were able to manipulate these loads on a very short-term basis to help the power grid out with some of its very short-term needs, say in three- or four- or five-minute increments. Those manipulations were really cheap because we weren’t really changing anybody’s temperature of their house much in three minutes, but we were helping the power grid a lot.”
On average, consumers kept about $150 at the end, about 10 percent of their electric bill.
Pratt says he knows of similar demonstration projects being run now in various pockets of the country, and he expects setups like GridWise to be fairly commonplace in the power-grid world in 10 years or so. “Eventually, these projects will start moving from the hundreds, like we did, to tens of thousands of customers,” he says.
The primary hurdles right now are regulatory, not technical, and that will change slowly since power companies and PUCs are conservative by nature.
“A major part of our activities now is reaching out to state and federal regulators to make sure they understand the potential and the practicality of operating the grid this way,” Pratt says. “We also want to make sure that the utilities are able to receive a reward for investing in these kinds of approaches, as opposed to new power plants and substations.”
In fact, Pratt estimates that adopting an SOA-EDA market-based approach across the United States could result in huge savings in power-grid infrastructure. “We are going to build a half a trillion dollars of new generation, transmission and distribution facilities in the United States in 20 years just to meet the load growth of our population and economy,” he says. “And we can save at least 10 percent, maybe 20 per cent, of that investment with these distributed, Internet-type control approaches.”
That doesn’t mean utilities never will need to build another power plant or substation. “But if we can save 20 percent of half a trillion dollars and all we’re spending is on information technology, that’s cheap compared to iron, steel and concrete. It’s a big opportunity and we’re very excited about it,” Pratt says.