Over the past few years, “distributed intelligence” has become a catchphrase among advanced metering infrastructure companies like Itron, Silver Spring Networks and Landis+Gyr. The idea is to leverage advances in low-cost distributed memory and processing power, and the increasing sophistication of the networks that bind them, to turn what have been mostly proprietary-technology-based, purpose-built devices into internet-capable applications development platforms -- in other words, something more like a smartphone.
It’s a common-sense idea, but one that’s only slowly making its way into the utility landscape. In part, that’s due to the slow technology adoption curve for utilities, which can take years to move from lab tests to pilot projects to commercial deployments. The slowdown in smart meter deployments, at least in North America, has also shrunk the pool of potential customers and put financial pressure on advanced metering infrastructure (AMI) vendors.
Then there’s the question of bang for the buck. Smart meters need to last decades, and cost less than $250 or so apiece in the United States or Europe -- or $50 or less for China, India and other emerging markets. You could sell a half-dozen generations of cellphone, at the same prices, in the same time it takes one smart meter to depreciate. Distributed intelligence has to cut costs and create new revenue opportunities that justify the costs of deploying it in this context -- and so far, AMI vendors that have been testing this kind of technology haven’t yet revealed any hard numbers on this measure.
In the past two months, however, U.S. smart meter maker Itron has revealed some important information on its plans to bring its distributed intelligence solutions to market -- including a deadline for doing it. At last month’s Itron Utility Week conference in Los Angeles, CEO Philip Mezey set the second half of 2016 for the launch of OpenWay Riva, its Linux-based, Cisco-grid-router-connected AMI network, to the broader market.
That’s about when Itron expects to complete its first OpenWay Riva deployment on the island of Tonga. A much larger deployment with Brazilian utility Eletrobras, along with Telefonica and Siemens, is set to begin next year as well.
In Europe, meanwhile, Itron is working with Salzburg, Austria’s municipal utility and telecom company Kapsch to test its wireless-plus-powerline Adaptive Communications Technology (ACT). Separately, it's testing some of the analytics applications it plans to support on its next generation of meters. At this month’s European Utility Week, the company demonstrated three of them -- advanced theft detection, high-impedance “hot spot” detection, and a localized demand response and transformer load management module -- on the same meters it’s deploying in Salzburg, albeit in a showroom in Vienna, not in the field.
These aren’t new ideas, by any means. Smart meters are being tapped as voltage sensors, outage detectors, home area network hubs, and other uses beyond their core digital cash register function. Their data, meanwhile, is being mined for asset management, theft detection, customer behavior analysis and load forecasting, and appliance-by-appliance energy disaggregation, to name a few analytics uses.
But if Itron delivers next year on what it’s demonstrating today, it’s going to be able to do these kinds of jobs much faster and more accurately than today’s AMI systems, using the applications-friendly, networked computing power in meters themselves to do much of the analysis and decision-making. In some cases, it’s promising to do things that today’s smart meters aren’t really capable of doing at all, at least not at a cost that would justify the use case.
That’s certainly true of the “localized, intelligent demand response” application that Itron demoed in Europe. It’s a good case study for defining what distributed intelligence means to Itron, how different it is from the way smart meters work today, and what its competitors will have to contend with as they unveil their own efforts.
The building blocks of a smarter smart meter network
First of all, it’s important to highlight the distinctions between Itron’s OpenWay Riva meters and the previous breed of smart meter. Older meters, in simple terms, consist of a metering chipset and a communication chipset. While the comms chipsets often have excess computing capacity and memory to perform distributed computing tasks, they’re not necessarily built to do so.
Itron’s new meters, by contrast, contain a Linux computer, capable of running applications that can collect, analyze and share local data with their peers or the network at large, Tim Driscoll, Itron product line manager, said in an October presentation at Itron Utility Week. This data is available at one-second or better resolutions, which is important for solving complex, distributed problems that arise in an electricity network that runs on 60 cycles-per-second current, he said.
The network itself uses standards-based, IPv6-compliant wireless technology, as well as powerline carrier for hard-to-reach places like apartment highrises in Hong Kong, where Itron first piloted this technology. Each meter is also capable of running multiple applications in concert with other local distributed intelligence nodes, without the need for instructions to come from a central command point, as in a client-server model, or from a far-off grid control room.
“Meters talk to each other, figure things out, and solve problems directly,” he said. In other words, where most distributed data analytics models deliver their outcome to a back-office system, which then responds with signals back to the devices at the edge, local analytics delivers its outcome to the edge devices themselves.
These edge devices can be other meters, or local grid routers, or transformer monitors, voltage sensors, or substation SCADA systems, working back up the distribution grid. Or they could be thermostats, solar inverters or electric-vehicle chargers, to name a few of the behind-the-meter devices that Itron is working on with its Riva development partners.
Importantly, they’re all running on the same network, he said. That goes a long way toward eliminating the kinds of challenges that can arise in bridging multiple devices and networks, particularly if they need accurate, sub-second data to make decisions. Using IPv6 also allows network traffic prioritization, multicast as a network protocol layer, and other features that help ensure that data gets where it needs to go in time.
Because each meter can run multiple applications simultaneously, it can also “translate” between different communications protocols -- DNP3 for distribution grid devices, Modbus for inverters, Wi-Fi for smart thermostats, and the like. And once these devices are trusted to make decisions locally, they can do things that central control systems struggle to achieve -- such as integrated, localized demand response, he said.
Pulling grid-edge pieces into a real-time community
Itron’s definition of this application starts with transformer demand management -- that is, controlling how much power homes and businesses are using to keep their local transformer from exceeding certain limits. Distributed intelligence can start helping with that application in ways one might not expect -- like, for example, helping to know exactly where each meter is on the distribution grid, in relation to each transformer.
Today’s AMI networks know where they are in terms of their address -- but their location on the grid itself may or may not be correctly noted in the distribution grid management systems that utilities use to monitor transformer loading, among other things, he said.
Using voltage and current data collected in one-second increments, Itron’s meters coordinate with each other and with utility distribution management systems, and “are able to figure out themselves where they are on the distribution network,” he said. While meters don’t know much about their local transformer, utility back-office systems can provide them the data they need, such as loading capacity, he said.
At the same time, Itron’s meters can offer a lot more data about what’s happening on the demand side of the equation, he said. Traditionally, demand-response programs enlist lots of customers to agree to turn down air conditioning, shut off pool pumps, and otherwise reduce energy consumption on utility command, with little ability to collect real-time data on what actually happens when the call goes out.
More advanced two-way demand response technologies can verify and calculate load drops, or differentiate between different endpoints they’re communicating with, he said. If they’re integrated to utility systems that monitor transformer loading, they could be asked to coordinate their demand response signals to specific parts of the grid to help alleviate local congestion or overloads, he said.
But Itron’s system combines these two features of transformer monitoring and behind-the-meter demand response in a single package. Or, to be more precise, it combines them into what Driscoll called “hives” -- groups of meters associated by their relative connections with their portion of the distribution grid.
Each hive of, say, 100 meters is constantly measuring each individual home or business, to assess how much power it’s using, and how much it has to reduce. But it’s also tracking its average usage, and how much total flexibility the entire hive has to reduce load and help out that local transformer. Then, when the demand-response call comes, the hive decides which loads to reduce to meet its quota, does it, and reports the results to the central control system.
It’s not that a utility couldn’t accomplish something similar without distributed intelligence. It could integrate its own transformer load monitoring systems, and its separate demand response systems, with the AMI system that keeps in contact with individual meters, and group those meters into appropriate categories. It could write the algorithms to manage the interplay of different behind-the-meter loads against what the transformer needs, coordinating each system with the other to use the most recently available data from each to cross-check the other’s performance.
But a distributed intelligence solution offers a far more elegant solution to the problem, Driscoll said -- particularly when the systems being integrated start to multiply. For example, Itron’s new meters can also sense whether power is flowing from substation to meter, as it almost always does, or from rooftop solar-equipped homes back toward the substation, as solar-rich countries like Germany are seeing today, he said.
If this distributed generation starts to affect local grid voltages, these meters sense that, and inform centrally controlled volt-VAR optimization (VVO) systems -- a task to which smart meters are already being put today. It could also tap locally available loads, or reactive power-capable devices like smart inverters, to help solve voltage problems, or ask household loads to increase their energy consumption to suck up excess solar power.
Quantifying the (extra) value in distributed intelligence
So, what is this all worth -- and when are utilities going to start turning on the applications?
“We’re working through this with our customers, what the additional value would be for implementing this,” Pieter Coetzee, senior director of business development, said in an interview this month. Based on a basket of early-stage applications like outage detection and restoration, energy-theft detection and load monitoring, “we’ve started sharing some preliminary numbers with customers,” with paybacks “in the order of north of 10 euros per year, per device, that you’re unlocking. Take that over a 15-year lifecycle,” for a device that costs about 200 euros, “and that’s significant.”
That's compared to a standard AMI network, not to having no smart meters at all, Tim Wolf, Itron’s smart grid solutions marketing director, noted. “As sort of a first pass, using our business modeling tools and experience, we’re seeing up to a 50 percent increase in utility benefits in some situations, when you factor in some of these distributed analytic applications.”
Benefits vary by application, depending on how closely they can be aided by distributed intelligence. Diversion detection, also known as theft detection, may be an early leader, based on the company’s results. “We think the Riva approach, based on flow of current, versus the old methods” that use 15-minute or hourly meter data, “yields a 300 percent increase in accuracy of detection,” he said. Theft detection is one of the main drivers for AMI business cases in markets like Brazil and India, and can translate directly to dollars and cents saved.
For outage management, Itron’s methods can yield a roughly 50 percent improvement compared to how AMI-supported outage detection is done today, he said. That’s mainly through utilizing peer-to-peer communications, along with the battery power on Cisco’s grid routers, to reduce the outages missed because the message got stuck at a meter without power on its way to reporting back up to the main system. The value here is measured in improved outage restoration metrics, with a value in avoiding fines for missing targets, as well as improved customer satisfaction.
Other applications yield benefits that are harder to quantify in monetary terms, he said. “For high impedance detection, there are questions about how you put a value on safety,” he said. But then again, detecting downed circuits that haven’t tripped alarms, but are still a hazard to anyone in the vicinity, is “something the utility could not feasibly or effectively do on the low-voltage network,” absent some kind of network of distributed sensors that can analyze grid power to find them, he said.
It’s likely that utilities are going to start with data analytics applications. Turning over the actual control of grid devices to that distributed intelligence, as with its intelligent demand response application, remains a big step into the unknown.
“We’re all very eager to get a field demonstration of these capabilities up and running,” said Ty Roberts, Itron’s marketing director for Europe, Middle East and Africa. “In some cases, where peer-to-peer communications is a central component, it’s going to take time for utilities to trust that our technology can actually do this.”
