22 Jul 2013

Real-time, big-data-enabled, demand reductions becoming THE go-to smart grid app

Written by Jim Pierobon

Of all the forward-thinking at this month’s 10th annual Town Meeting on Demand Response and Smart Grid in Washington, DC and similar conferences in the U.S., one new application is appears curiously off the main radar screen.

I’m talking about software that utilities can use to automatically program large-scale demand reductions in real time by harnessing lots of big data. One company driving such advancements: AutoGrid in Silicon Valley.

Thus far, municipal utilities in Austin, Palo Alto and Sacramento, along with Oklahoma Gas + Electric and Southern California Edison are testing AutoGrid’s algorithm. If it stands up to these and other tests, every utility commissioner with authority over rate recovery for new power plants, including those at NARUC’s summer meetings this week in Denver, should first ask: how the utility can meet demand with large-scale programmed demand reductions before they authorize higher rates.

Demand response may be the most exciting and fast-changing space in the energy sphere. What used to be a ‘cottage’ industry is now a thriving sector in this part of the smart grid space. Most utilities with demand-response programs rely on third-party providers such as Comverge and EnerNOC; a few are said to be developing their own systems. These programs can pay back big time during heat waves such as the one that gripped the Eastern half of the U.S. last week.

AutoGrid appears to be unique because it enables utilities to manage large-scale, direct load-controls in-house and do so virtually instantly. Most utilities may not yet have the technical expertise to deploy AutoGrid’s and similar software on their own. Installing smart meters is one thing. Deploying them and then getting their arms around the exponential growth in data is something else altogether.

Indeed, there seems to be an element of risk involved if a computer reduces power to a batch of customers mistakenly. These and other demand response programs may expose wholesale power markets to potential manipulation. But if AutoGrid’s software is accurate, reliable and easy enough to deploy, the potential is huge, far beyond what “demand response’ means to most industry pros and regulators today.

Some venture capitalists see ‘green’ in the way AutoGrid is moving to harness big data. They’ve provided $7 million in two rounds of funding since 2011. AutoGrid claims it has about $20 million in its business development pipeline.

image002

Here’s one way to envision the challenge the utility industry faces in leveraging the exponential growth in data available from smart meters. From one meter reading a month   . . . . .

image002

. . . . to almost 3,000 readings a month. CREDIT: AutoGrid

The company’s sales pitch includes a 1950-era utility employee, left, taking one reading a month followed by a graphic trying to drive home the enormous amounts of data utilities have access to — about 96 data uploads a day — with the deployment of smart meters.

Various sources estimate the number of smart meters installed in homes and businesses throughout the U.S. to be about 65 million and growing every month.

The AutoGrid algorithm analyzes all this data to predict future demand for electricity customer-by-customer. In an interview with The Energy Fix, Founder and CEO Amit Narayan said AutoGrid is is capable of generating demand forecasts for more than 1 million customer accounts every 10 minutes. The forecasts can be aggregated by class and zip code.

“This is really the big promise of the smart grid,” he said. “All this data sits in a database but it’s not being used. We’re trying to unlock the value.”

AutoGrid and other software are built on a standard known as “Open ADR” for a transparent means of automating demand responses between electricity providers and grid operators via the Internet using a common computer language.  It has gained widespread support throughout the power industry and is subject to engagement rules by the grid operators such as PJM Interconnection and the California Independent System Operator.

Narayan explained the research leading to the founding of AutoGrid was his first foray into energy; this after he directed smart grid simulation research at Stanford University. Much of his previous work focused on developing software to design semiconductor chips. Out of that grew one of the more noteworthy successes he’s associated with: developing 80 percent of the chips used for Apple’s first iPhone.

The senior team at AutoGrid includes two ‘alums’ from Pacific Gas & Electric, Chris Knudsen and Andrew Tang. Interesting (to me anyway) how many apparently innovative professionals earned early stripes at PG&E which has struggled at times in deploying smart meters. That tells me there will always be value in learning from a company’s shortsightedness.

Implementing AutoGrid’s software at a large commercial or industrial facility is relatively straightforward. In a recent trial by Palo Alto, customers with 26 megawatts of electricity load — representing 14% of the city’s peak load — were set up within 30 days. On average, the city’s system has dropped 1.2 megawatts per event, saving 3.5 megawatt hours per event.

The cost to set up a large customer is $70,000 per facility. Scaling up the business could drop that cost to $500 per facility, Narayan said.

He said AutoGrid is working on software to integrate wind and solar power systems. If that software gets traction, it would answer one of the strongest arguments against the intermittency of large-scale renewable energy systems.

Here is an excellent discussion at a recent Bloomberg conference featuring Narayan on how deploying big data quickly is becoming the ‘next big thing’ on the smart grid horizon.

Leave a Reply