We’ve long heard rumblings from Silver Spring Networks (SSNI) about a big project in the works, one that puts the company's millions of connected smart meters, its IP-based networks, and its cloud computing architecture to use in a way not seen before in the smart grid space.

On Monday, the Redwood City, Calif.-based company revealed how it will put that vision into practice. It’s called the SilverLink Sensor Network, a reimagining of the role of smart meters, grid sensors, home energy monitors and other devices at the grid edge. Silver Spring sees these not as discrete devices, sending data through disparate channels up to constrained utility back-office systems, but as virtual nodes of a distributed intelligence platform that are able to collect, process and trade data between themselves a bit like web servers do today.

Consider it an attempt to replace old-school field and enterprise service bus architectures, extract, transform and load (ETL) back-office data integration methods, and other hallmarks of utility IT architectures with a faster, cheaper and more agile way to link real-time data collection and analytics to utility business processes. Indeed, Silver Spring claims it could allow utilities to “create breakthrough business applications at ten times the speed and one-tenth the cost” of traditional data warehouse and analytics.  

That’s a big promise, and Silver Spring has a way to go to prove them out. Monday’s launch includes some core applications like meter and network operations management from Silver Spring’s existing cloud services for utilities, that leverage its new distributed computing benefits. But the company is promising to add new grid-facing, business-facing and customer-facing applications built in-house and by utility partners, with Oklahoma Gas & Electric and Pacific Gas & Electric named as two early adopters.

It’s also promising to provide interfaces to smart meters, grid equipment and devices from other vendors. “For the first time in our history, we have a product that [can be used outside of] a Silver Spring network,” Anil Gadre, Silver Spring executive vice president of products, said in an interview. That could extend to connecting meters from other vendors, or to smart thermostats, networked streetlights, and a myriad of devices that lie in the nebulous category of the “internet of things,” he said.

And then there are the third-party applications that an open-standards-based platform can enable. Monday’s launch names seven of them, including three existing Silver Spring partners: data analytics startup AutoGrid, grid sensor startup Sentient Energy, and volt/VAR optimization provider DVI. It's also added NoSQL database vendor MongoDB, pointing to some hefty data analytics applications in the works.

Three more apps partners -- Bidgely, PlotWatt and Onzo -- provide energy disaggregation software that can tease out individual appliance energy usage from whole-home energy data. That’s the kind of functionality that might allow utilities to challenge third-party contenders like Nest and Google in delivering energy insight to utility customers, Gadre noted.

The overall goal is to create “a platform on which the customer, the utility, is going to begin the process of true transformation of the business,” he said. On this front, Silver Spring is sure to face competition from rivals seeking to bring distributed intelligence to smart grid endpoints. One big one to watch is Cisco and its IPv6 grid networking platform for smart meters, grid devices and “smart city” nodes like parking spaces and streetlights -- Cisco will no doubt be moving to open that network to broader uses. Cloud-based smart meter data analytics from eMeter and Oracle could also challenge Silver Spring. 

And then, of course, there are Silver Spring's internal needs to take into account. The company is under pressure to grow its share of software and services business, both to add value to its smart meter networking deployments, and create ongoing revenues to buffer the ups and downs of AMI contract wins in a challenging market. In broader terms, utilities will be increasingly shifting their focus from grid hardware deployments to the data analytics tools they need to optimize their effectiveness over the remainder of the decade, making analytics the new frontier for grid vendors of all stripes.

So how does Silver Spring plan to break the mold on smart grid IT architectures, increase the value of its networks for existing and future customers, and keep ahead of the competition?

The Smart Grid Network as Computing Platform

First, imagine the smart grid network as “the first continuous connection between the home and utility,” as Gadre put it. If that network represents a superhighway, “we forgot to build the off-ramps to the inside IT processes that need all this data,” he said.

That’s the problem that utilities face when they try to make use of traditional IT systems to process the floods of data that the smart grid delivers. AMI data represents about three-quarters of this new flood, and the costs and complexities of integrating it into back-office systems has kept it from being used properly -- a fact that’s borne out by a host of surveys on the topic.

In the meantime, “A meter is out there, and it’s actually a very smart tool, but it’s being very poorly leveraged,” said Scott Young, Silver Spring’s senior director of software applications and analytics. With the meter’s computing power, as well as the Silver Spring NIC (network interface controller) card inside it, “let’s just virtualize it, let’s give it an address,” he said. “It lives on the network; let’s use it in ways that make the most sense.”

The same goes for other devices linked to that network, while some are “smarter” than others, all have some processing and data capabilities, he said. And critically, unlike most traditional back-office utility systems that batch-process the data they’ve collected in intervals of hours to days, these distributed devices can compute at the speed of the network, he said -- as long as they’re properly related to one another. 

That’s where the concept of virtualized sensors, or software-defined sensors, comes in. “The most important thing that’s happening with the SilverLink sensor network is that we’re virtualizing every device. An application does not need to be pinging every meter, or every device -- it only needs to talk to the virtual sensor,” he said.


SilverLink Sensor Network takes three critical steps to make this possible, according to Alex Zheng, senior product manager. “First, we make it easier to get the data off the network,” by setting up each virtualized end node to collect, interpret, filter and deliver the specific types of data needed for each particular application. That’s done via web programming techniques known as representational state transfer (REST), and Silver Spring’s new platform calls itself “RESTful,” in that it meets all the requirements of a REST-based architecture

“Second, we standardize the interface to all this data via our API,” or application programming interface, which allows each application built for the network to make use of these commonly defined virtual sensors in consistent ways. Finally, we make all that infrastructure elastic,” by pulling in Silver Spring’s cloud computing resources, he said. Because that operates independently of the utility back-office IT systems that Silver Spring’s AMI network feeds meter data to, it isn’t constrained by those systems’ limitations.

Silver Spring does face limitations is in the bandwidth and latency of its wireless mesh, which hovers at around 250 kilobits per second, and requires data packets to take multiple “hops” between devices as it winds its way up to the collection nodes that send it back to the cloud. But Zheng said that, beyond the core meter-to-cash data collection it was built for, “typically, we’ve found that the mesh network has a lot of excess capacity that can be used for different applications.”

As for latency, “we’ve found that 95 percent will take eight seconds or less” to make that trip from endpoint to collection point, “similar to what you’d expect to get through other types of networks.” Silver Spring promises it can make this all happen without affecting the quality of service (QOS) agreements it has on delivering its core meter data, of course, Gadre noted.

Valuing the Power of Grid Edge Data Awareness and Analysis

How does all of this translate into faster and more flexible computing and data analytics? For his first example, “Let’s give the customer an itemized energy bill,” Young said, using the energy disaggregation of the kind SilverLink Sensor Network partners Bidgely, PlotWatt and Onzo are doing.

These systems use meter data, which measures total customer electricity usage, to analyze and dissect how much power is being used by individual systems in the home or business. “The ability to disaggregate energy and find out what appliances are behind this load, the value is tremendous, even on a small opt-in deployment basis,” Young said.

But these systems rely on lots of data, as well as data that comes quickly enough to be turned around into close to real-time information for customers. “What happens if, in order to give me an itemized bill, I need one-minute reads? I know the meter is capable of it,” he said. But for a utility, “there’s a million reasons why a utility can’t, most of which are because of the batch process mode” that involves fifteen-minute to hourly interval meter reads being fed to a meter data management (MDM) platform that processes the data about once a day.

To revamp that back-office system to push one-minute data out the end, “we’re going to have to spend millions of dollars, it’s going to be 60 times the data we’re now collecting,” he said. But the SilverLink Sensor Network can collect that one-minute interval data via its virtualized sensor in the meter, compress it and move it through the network to a cloud-hosted server that runs the deeper analytics for utilities to use, or even send it back to deliver close to real-time information to customers.

Similar techniques could bring much faster and richer data to fine-tune demand response and energy efficiency programs, he said. On that front, Oklahoma Gas & Electric is “our poster child for pretty near real-time connection to the customer,” with a Silver Spring-enabled set of smart thermostats, time-of-use pricing and demand response programs that have won accolades in the industry. Silver Spring’s existing line of Customer IQ services for OG&E won’t change, but “the sensors underlying it will allow us to do those thing much more quickly,” he said.

Beyond that, “the more data we get back into that system, the more contextualized that data becomes, the better we get at forecasting where that need will be.” That’s important for helping partner AutoGrid perform its predictive data analytics, which require frequent and rich data inputs to work to optimal effectiveness. In California, big customer PG&E is also doing some R&D projects into how it could put some of these customer-facing applications to use, he said, though he declined to disclose more details.

Moving up the distribution grid, the same techniques could be used to deliver specific and timely data to theft and loss detection systems that pinpoint where electricity is being wasted or stolen, outage management systems to pinpoint faults in the local grid, or conservation voltage reduction (CVR) systems that manage grid power for energy efficiency or peak demand reduction, Zheng said. These aren’t new concepts for using AMI data, but pushing the computing out to the edge could make these tasks a lot faster and simpler to deploy.

But perhaps the most important measure of value lies in how SilverLink Sensor Network could lower the cost and time it takes to merge smart grid data with new applications. Silver Spring’s partners find that “when they go out and do a project, about 70 percent to sometimes 80 percent of it is how they get the data,” he said. “If you can standardize the data that’s coming of the network in [an] API that’s very easy for people to integrate, […] you’ve eliminated a huge upfront chunk of the cost to do it.”

“One of the key issues going ahead with this is to expand that ability to innovate, to decrease the amount of time it takes, and to increase the ability of partners to integrate,” he said. Just how quickly this “apps store” for utilities grows, and how well the platform performs on delivering them, will be important measures of success for this vision.

Tags: applications, apps, autogrid, bidgely, big data, cisco, data analytics, demand response, distributed intelligence, energy disaggregation, grid edge, grid platform, ip, oklahoma gas & electric, onzo