How non-intrusive load monitoring can enhance smart meters, save energy, and save money
There are now over 50 million smart meters deployed across the United States, with millions more planned for installation by ComEd in Illinois, by Consumers Energy in Michigan and many other utilities domestically and internationally. Smart meters measure and transmit customer load profiles at hourly or sub-hourly increments—measuring a customer’s apartment, single-family house, or small business—and represent a step forward from analog meters that were read manually once each month.
Smart meters importantly enable things like time-of-use pricing, but despite their name, smart meters have not yet proven intelligent enough to give customers sufficiently actionable information about the devices and systems they touch in their daily lives in a manner that interests them. As a result, many of the operational benefits of smart meters have largely accrued to utilities, not customers.
Enter non-intrusive load monitoring
A number of companies are developing increasingly more-accurate and transparent solutions to this problem to enhance the value of smart meter data for consumers, collectively known as non-intrusive load monitoring (NILM). NILM is analogous to an always-on electrocardiogram (EKG) for your home or business. It dissects electric meter load curves to “see” and diagnose the performance of the appliances and systems that use electricity that are embedded in the signal. In a sense, each major electricity-using appliance has a distinct “signature” that can be extracted from the whole-home load curve, like a recognizable heartbeat. By parsing out individual components such as air conditioners, hot water heaters, dishwashers, refrigerators, electric vehicles, and other loads, NILM disaggregates smart meter data (e.g., load) into its respective pieces.
NILM provides granular content customized for individual households to induce greater energy savings. Granular NILM content, if coupled with the right messaging channel and timing, could deliver on the promise of customer interfaces that motivate better customer responses from energy data. If we have learned anything from the brief history of Google PowerMeter, Microsoft Hohm, and dozens of still commercially available consumer portals that are based on the premise of just monitoring energy use (and who have failed to engage the mainstream consumer), it is that energy-efficiency stakeholders are not wired to think about what makes consumers tick. NILM expands customer benefits beyond energy use monitoring, and into appliance health notifications. In other words, it bridges from merely monitoring energy use passively to doing something about it actively.
For example, by identifying aberrations in appliance operation signatures, NILM enables virtual monitoring of potentially billions of devices behind customers’ smart meters that represent the majority of residential electric loads. The same systems could notify customers when their devices experience problems before they result in appliance or system failures. Because NILM leverages meter data, the customer does not have to buy a new Internet of Things (IoT)-enabled device that has an embedded communications chip and software package, offering similar notification capability.
NILM going mainstream
NILM is not new. However, it has evolved from an academic pursuit to recently launched commercial products now gaining traction in the market. Companies like EEme, Bidgely, Intel, C3 Energy, and others are racing to develop and validate the most accurate algorithm for software solutions that interface between customers and other actors in the IoT value chain. A barrier to commercial launch, up to recently, had been the absence of validation studies that used a robust set of ground-truth data proving the algorithms worked. Previous validation studies were conducted by EPRI and other research organizations that were not publicly available, and generally limited to 10–15 mock homes and time spans of 1–2 months.
Historically, NILM’s low accuracy, combined with the lack of publically available ground-truth data, has been a barrier to commercialization. However, both the accuracy of NILM algorithms, along with the availability of ground-truth data, is proving the technology merits a fresh look by utilities and third-party technology companies. Academic studies have previously set the benchmark for NILM accuracy at 55 percent, not much better than a coin toss. The accuracy of NILM algorithms has improved, now hovering around a much more impressive (and useful) 70 percent.
And the process is improving further and becoming increasingly more transparent, based on a March 2015 public test from EEme and Pecan Street Inc., which employed previously unavailable ground-truth circuit-level data that encompassed 270 homes and 12 months making it the most comprehensive of its kind. That transparency and rigor is crucial, since many in the industry are skeptical of load disaggregation’s value.
Unlocking customer energy savings
The grid of the future will be centered on the customer, enabling customers to understand and manage their energy use more efficiently. Personalized, transparent, and actionable data availability to customers and to the marketplace is a key factor enabling that transition.
Yet to date, customer engagement in the electric industry has not exactly been captivating. Many customers autopay their utility bill and thus never see it. We’re billed for energy we already consumed last month. And traditional, simple meters mean that we “see” little more than a lump sum of kWh consumption month by month. The result is the “standard” utility bill as many of us have known it:
Although customers intuitively can deduce greater energy consumption in hotter months, this bill doesn’t tell customers precisely what their biggest energy demands are and how much of that demand they draw at the most expensive times (which would be very helpful to know to manage customer bills and determine if they wanted to choose a time-of-use rate, for example).
Smart meters open up far greater insights, especially when paired with load disaggregation. Customers who consent to feed their smart meter data to an energy monitor app like PlotWatt (say, via Green Button Connect) might receive far more interesting information about their electricity use:
This information begins to tell a story about how a customer’s consumption correlates with the components of our consumption (specific appliances), the associated rhythm of our daily lives, and therefore the choices we can make to better manage them.
Going further, NILM allows for more-innovative visual displays of quantitative information, parses out the components of consumption, and conveys information about the performance of particular appliances. For example, the weekly home energy report from EEme similarly separates electrical loads graphically, provides an observation, and therefore recommends the customer inspect their hot water heater, and links to a real-time load profile.
NILM-enabled personalized feedback has been described as the “holy grail” of energy efficiency, and yields the greatest percentage of customer responses and energy savings. Dr. Carrie Armel at Stanford and Bidgely CEO Abhay Gupta led the 2013 benchmark study of the value of NILM, finding that appliance-level feedback, augmented with personalized recommendations, achieved the greatest percentage (e.g., >12.0 percent) of annual energy savings, in comparison to indirect feedback (after consumption occurs), as well as real-time direct feedback.
Source: K. Carrie Armel, Gupta, Shrimali, Altert, January 2013, “Is disaggregation the holy grail of energy efficiency? The case of electricity” Energy Policy, V. 52.
The study suggests that while time-variant pricing is an important component of achieving customer benefits associated with smart meter deployments, greater savings are enabled through direct, personalized feedback that accurate NILM solutions can uniquely provide.
Unlocking utility opportunities
As the meter owner and installer, utilities are a key enabling stakeholder in NILM. They’re also poised to seize opportunities inherent in NILM. For example, utilities can utilize NILM to identify customer loads of interest and improve targeted marketing efforts. Utilities are used to forecasting loads at an aggregate resolution (e.g., substation). With NILM they can forecast individual end-use-level customer loads in a bottom-up approach, and thus procure, transmit, and distribute power more intelligently and efficiently.
In the case of plug-in electric vehicles, utilities rarely know when and where a customer has adopted a plug-in EV absent customer notification. However, utilities may be interested to track that customer to avoid service interruptions in the unlikely event a distribution system upgrade is needed due to the increased load, or might target specialized rate offerings to attract that customer, as opposed to wasting money on non-EV owners. Rate designs themselves could adjust to personalized, appliance-level load profiles, as opposed to the clunky “one-size fits all” that tend to benefit some and penalize others for behaviors that are outside the assumed profile.
Further, utilities can leverage NILM to improve on existing methods such as interconnection and tools such as PV Watts to evaluate solar rooftop panel performance on customer homes. The impact of distributed generation on grid stability and security can be addressed by accounting for residential solar power through NILM. In turn utilities can also identify strategic locations for solar PV that can alleviate the load burden in more sensitive parts of the grid. NILM can thus empower utilities to guide and control distributed generation both in a consumer- and grid-friendly way.
The role for third-party technology companies
NILM has the potential to be a gold mine for IoT companies pursuing new customers, as well as for existing appliance manufacturers’ customer retention strategies. For example, personalized energy-savings recommendations are already available to and benefit owners of communicating devices, such as Nest thermostats. For customers that haven’t purchased an advanced thermostat, NILM will help third parties identify good leads that would benefit from a learning thermostat, e.g., by helping to identify those customers wasting energy to cool spaces when and where it isn’t needed. In addition, because certain appliance load signatures are identifiable through NILM, a company like Whirlpool can find existing customers through meter data mining before the customer needs to replace or upgrade their appliance; they can send maintenance or in-brand replacement coupons, retaining the customer.
In reality, few residential customers actually care about energy savings, spending less than six minutes a year thinking about energy use. IoT-enabled home security offerings from AT&T, Lowes, Vivint, and others indicate that customers ultimately care about 1) safety, 2) security, and 3) comfort. Remote appliance monitoring and management is therefore a more attractive value proposition, by allowing family members and third-party market actors to monitor the “health” of HVAC and related hardware that provides their loved ones with security and comfort.
Concerns and opportunities
To be fair, in many places the value of smart meter data has not yet been realized because of regulations that prevented initial market access to smart meter data, as well as policies that prevent default time-variant pricing, which has potential to yield tremendous savings.
However, NILM has the potential to significantly enhance the value of existing meter assets for customers, and allows everyone to leap forward and achieve efficiencies enabled by machine-to-machine communication, without the added costs of adopting IoT-enabled devices. As more people adopt IoT devices, NILM can complement these services, but since a fully smart home might be a ways away for many Americans, it turns out the meter sitting outside their home has the potential to accelerate that transition.
Coauthor Enes Hosgor, Ph.D., is the founder and CEO of EEme, LLC.
First image courtesy of Shutterstock.