Big Data & Analytics: A Widespread Phenomenon
Smart meters and other intelligent field assets are allowing utilities and others in the electricity supply chain to streamline operations and cut costs. Eliminating manual meter reads alone is collectively saving billions of dollars annually and reducing millions of truck rolls.

And this is just the tip of the iceberg. Advance Meter Infrastructure (AMI) is also building an incredibly valuable asset that will prove even more valuable in the long run – data about customers, behavior, and ongoing operations.

‘Data’ can be a difficult asset to value, especially when the proper analytics systems are not in place to harvest the full value of all data streams. But take a look at a company like Google. The search engine dominates information management because it provides quicker and more accurate results to queries than competing technologies. Its data centers and algorithms continue to improve because accumulated queries allow it to deliver more precise results (with fewer underlying processor cycles) the next time around.

Or consider Netflix. The entertainment service has mined data about the behavior of its viewers to become a successful TV producer. No more guessing what the audience wants:

• Netflix can intelligently anticipate it with a relatively high degree of accuracy.
• People are predictable. It’s one of the most important, but one of the most underappreciated, insights of the Internet era.

Similarly, Amazon has developed recommendation engines, optimized data centers and the logistical framework that have allowed it to become the most dynamic retailer in the world.

Customers can navigate the site, and be prodded into impulse buys with personally targeted discounts, because Amazon has figured out how to guide them by studying their own past behavior.

Amazon has even taken it one step further by parlaying its expertise in data management into Amazon Web Services, one of the largest cloud service providers in the world. It’s a remarkable transformation: if you asked someone five years ago which company was destined to bring cloud platforms to mainstream businesses, they probably would have guessed IBM or Oracle, not the people who invented Cyber Monday. Data analysis is a competitive advantage as well as the foundation of their operations.

Utilities are sitting on the verge of a similar transformation. Historically, utilities consumed a relatively small amount of data for their size. Now, the proliferation of smart devices has created a veritable tidal wave of digitized information. A typical smart meter is serving up 2,880 meter reads a month, versus the one per month delivered by an analog meter. By 2020, the 980 million smart meters around the world will generate 8241 petabytes of data a year or more than 68,000 times more data than that is currently held in the Library of Congress. Building management systems for office buildings will generate around 100 gigabytes of information a year.

This data, if mined and analyzed in constructive ways, will ultimately give utilities a way to see patterns amid chaos. They will be able to accurately determine power going to plug loads versus HVAC or lighting, visualize and calibrate consumption rhythms throughout the day or predict the behavior of classes of customers or specific individuals.

With these forecasts in place, the next step is actually comparatively easy: harnessing these predictive abilities to fine-tune transmission, distribution and consumption. Demand response is already being transformed. Conventional demand response services rely heavily on planning, phone calls and advance organization. Demand response events are truly ‘events’ and occur only a few times a year.

Data-based systems turn demand response into demand management. Power savings can be harvested from hundreds of thousands of customers, not just a few. Securing participation of specific individuals becomes less crucial: if one customer can’t participate, many others can. Events can be organized quickly through a network and, because events can be scheduled more frequently, customers become acclimated to the service and increase the frequency and depth of their participation. Demand response goes from being an unusual event that requires a perceived sacrifice to a monthly, or even daily, occurrence with a benefit that is articulated on in dollars and cents on monthly bills.

Similarly, industrial customers will begin to adopt cloud-based systems to help control demand charges. Demand charges can account for 30 percent of a large power user’s bill. By employing intelligent automation, large power users can turn down less essential power consumption (like daytime lighting), maintain production flows and avoid excessive peaks. Without data, large power users can only guess what their power demands might be: data effectively eliminates risk by tightly defining probable outcomes.

Data can also be used to throttle power theft. The World Bank estimates that $85 billion in power gets stolen every year. In emerging nations, the problem is a never-ending crisis: approximately 30 percent of the electricity gets stolen in India, leading to chronic outages, lower productivity and higher rates. But it’s also a problem in the U.S. with $6 billion alone being siphoned off by illegal marijuana growing operations, among other activities.

Unfortunately, it is also extremely challenging to stop. Electrons don’t have serial numbers. There is no forced entry – the thieves live or work in the same location where they steal power – and police do not have the resources to dedicate to it. Utilities, rightly, also do not want to send employees to illegal drug operations.

With data, utilities can identify anomalies; compare actual consumption with billing rates and other parameters to hone in on probable suspects quickly and at a safe distance. The benefits and the results that we will see from data-based services will cause the Berlin Wall that separates utility operations, or OT, from IT to crumble sooner than many expect.

Time-of-use (TOU) programs, buffering for solar and wind farms, grid balancing will all suddenly become more feasible through analytics. Rather than just know supply, utilities will have their fingers on the pulse of demand. Incentive programs can be sculpted for particular customer segments and designed for maximum participation. Forecasting has always been the Achilles’ heel of TOU or efficiency programs. Utilities and regulators put the programs together, but then struggle to find participants. In many jurisdictions, incentive payments remain invisible. Analytics succeed because they eliminate the guesswork.

The power industry will never be a rapid technology adopter and for very good reasons. Utilities are the ultimate just-in-time industry. They cannot go offline over the weekend to conduct routine maintenance like a gaming site. Likewise, customers aren’t going to shrug at outages, chalking it up to the cost of adopting new technologies: they will view it as an inexcusable failure.

But the economic arguments in favor of employing data – and, more importantly the value in the insights apparent as utilities and customers become immersed in these systems – will become overwhelming. With this technology, utilities can shift from being perpetually mired in reactive mode, i.e. the kind of company customers only remember when something goes wrong. They can provide new services, lower their own operating expenses, reduce outages and increase customer satisfaction at the same time.

In this case, the future is easy to predict.
 

About the author

Dr. Amit Narayan is the Founder and CEO of AutoGrid, Inc. From 2010 to 2012, he was the Director of Smart Grid Research in Modeling & Simulation at Stanford University, where he continues to lead an interdisciplinary project related to modeling, optimization and control of the electricity grid and associated electricity markets. Prior to founding AutoGrid, he was the Vice President of Products at any Magma Design Automation. Dr. Narayan also founded Berkeley Design Automation, Inc. (BDA), a venture-backed company in analog and radio-frequency semiconductor design software. In 2006, Dr. Narayan’s received the EDN’s ‘Innovation for the Year’ award. Dr. Narayan received his B. Tech. in Electrical Engineering from Indian Institute of Technology at Kanpur and Ph.D. from University of California at Berkeley. He has published over 25 papers about design automation, holds seven U.S. patents and is an active advisor to several startup companies in the Bay Area of San Francisco.