Submetering II: In the Know with Electrical Flow
 

Submetering II:
In the Know with Electrical FlowWhich is more convenient? Submetering a house by stringing little electric meters to each circuit, or hooking up a new electric meter that uses math to "disaggregate" the electric use and determine which appliances are being turned on and off? A system originally developed at MIT in the 1980s has matured into an electric meter that can disaggregate electricity flows. The meter, developed by the Electric Power Research Institute (EPRI) and Telog Instruments, allows collection of submetering data without invading consumers' homes. According to Mark Malmandier, product manager at Telog, "The meter can also spot malfunctioning appliances, helping prevent major electricity drains, and diagnose high bills."

The Nonintrusive Appliance Load Monitoring System (NIALMS) works by identifying loads as they turn on and off, and matching them against a database of electric profiles-the same method being used in Japan for gas submetering (see "Submetering I: No Guessing with Gas," p.6).

Electric submetering must differentiate among the nearly limitless number of home electric appliances. Fortunately, everything that uses AC power has an electric profile in both watts and reactive volt-amps (VARs). Each appliance draws current. It uses up some power, measured in watts. It also draws some power that it doesn't use up, which is used to create electromagnetic fields-in motors and fluorescent light ballasts, for example. This "reactive" power is measured in VARs. Normal utility meters measure watts used over time to obtain kilowatt hours (kWh) for billing purposes. NIALMS also measures VARs.

Figure 1. Appliance transition clusters. The NIALMS system recognizes changes in household electric flow, and associates those changes with specific appliances. For example, increases in watts and reactive volt-amps that fall into the "heat pump" oval are identified as a heat pump turning on. Matching decreases are recorded as a heat pump turning off.

Each appliance's power and reactance can be graphed on a scatter chart (see ); an appliance's unique profile is known as its signature. As appliances turn on or off, NIALMS notices the change and compares the new signature with those in its preprogrammed database. It is programmed to recognize 140 common appliances.

NIALMS records the on-off information in memory and periodically sends the data to a central computer at the utility. NIALMS can communicate over phone lines or power lines, or via hand-held data collectors used by meter readers. Most units in field tests have used shared phone lines, typically using the resident's phone line to automatically dial the utility for periodic downloads. The call-in schedule can be dictated by the utility or selected by the resident.

At this point, the system is about 85% accurate compared to direct metering. Large appliances, such as heat pumps and refrigerators, are detected accurately over 90% of the time. While refrigerators are generally detected properly, their 400W defrost cycles are often misidentified. Some variable-power appliances, which have multiple signatures, are missed entirely.

The meters are programmed to ignore loads below a certain wattage, dropping them into an "other" category. In the default setting, the meters do not submeter loads below 150W, so lighting is largely missed. Most utilities using the meters in tests have actually requested a higher minimum, so even fewer loads are monitored. The utilities generally prefer to focus on high-wattage appliances that cause spikes in electric demand, such as air conditioners, heaters, and clothes dryers. Once the utility has the data, the information can be used to monitor conservation strategies and to help resolve billing complaints.

Shown here are the interior of the NIALMS meter and a computer screen displaying some of the disaggregated data as it will appear at the utility.

As an additional benefit, NIALMS quickly spots major abnormal power uses. For example, in field tests, researchers have discovered faulty refrigerator compressors, a defective water bed heater, a broken air conditioner, and an electric heater positioned to start up whenever the air conditioning fan cycled on.

Despite its many features, the future of NIALMS is not assured. Traditional electric submetering requires disruption of the subject's house, but it costs much less. NIALMS is nonintrusive, but it is expected to retail for $1,200 per meter plus $15,000 for the utility-based server and its software (each server will be able to monitor 300 meters). Utilities will probably not pay these prices to replace all their basic meters (which cost about $150), but the information from NIALMS may be useful for marketing departments, for developing rate plans, and for resolving high bill complaints.

In the future, NIALMS will be programmed to recognize more appliances and it is expected to get better at recognizing appliances with changing signatures. Since NIALMS communicates over phone lines, it can eliminate human meter readers, as more and more utilities are looking to do. Finally, according to Malmandier, itemized billing may be in the works, since "some customers may want an electric bill that looks more like a phone bill."