Nonintrusive Appliance Load Monitoring

Contents:
  1. Technical Overview
  2. Multistate Appliances
  3. Development Status
  4. People to Contact
  5. References and Related Links
This page is was created by George Hart many years ago, and is now quite out of date.  I am no longer doing research in this area and can provide no more recent information, but here are links to two recent conferences, which should lead you to current information:

  http://nilmworkshop.org/
  http://www.ices.cmu.edu/psii/nilm/

Addendum, 2020: For recent developments in this area, please contact Stephen Makonin at Simon Frasier University (smakonin@sfu.ca)


Technical Overview

A Nonintrusive Appliance Load Monitor (NALM) is designed to monitor an electrical circuit that contains a number of devices (appliances) which switch on and off independently. By a sophisticated analysis of the current and voltage waveforms of the total load, the NALM estimates the number and nature of the individual loads, their individual energy consumption, and other relevant statistics such as time-of-day variations. No access to the individual components is necessary for installing sensors or making measurements. This can provide a very convenient and effective method of gathering load data compared to traditional means of placing sensors on each of the individual components of the load. The resulting end-use load data is extremely valuable to consumers, energy auditors, utilities, public policy makers, and appliance manufacturers, for a broad range of purposes. For example, a monitor placed outside a home can determine how much energy goes into each of the major appliances within the home.

In a utility application, a NALM connects with the total load using the standard revenue meter socket interface, as shown in the figure above. This permits very easy installation, removal, and maintenance compared with traditional intrusive load monitoring techniques that require ``submetering'' and interior wiring. The NALM monitors the total load, checking for certain ``signatures" which provide information about the activity of the appliances which constitute the load. For example, if the residence contains a refrigerator which consumes 250 W and 200 VAR, then a step increase of that characteristic size indicates that the refrigerator turned on, and a decrease of that size indicates the turn-off events. Other appliances have other characteristic signatures. After determining the exact on and off times from the signature events, any desired statistics, such as energy consumption vs. time of day or temperature, can be tabulated.

To appreciate how this works, consider this figure, which plots total (real) power consumption vs. time for a single-family home over a two-hour period. During this interval, the total load shows activity due to a refrigerator and a heater. Two different-sized step changes are clearly present, providing characteristic signatures of the refrigerator and the heater. The refrigerator cycles on and off three times, the heater six times. By measuring the total load ouside the home, it is not difficult to find these step changes and measure their size. Knowing the time of each on and off event, the total energy consumption of the refrigerator and the heater are easily determined. By also considering measurements of the total reactive power or harmonic current, along with the real power shown, changes in the resulting vector function of time would reveal even more information about the particular appliances.

Traditional load research instrumentation involves complex data-gathering hardware but simple software. A monitoring point at each appliance of interest and wires (or sometimes power-line carrier techniques) connecting each to a central data-gathering location provide separate data paths, so the software merely has to tabulate the data arriving over these separate hardware channels. The NALM approach reverses this balance, with simple hardware but complex software for signal processing and analysis. Only a single point in the circuit is instrumented, but mathematical algorithms must separate the measured load into separate components. In many load-monitoring applications, this is a very cost-effective tradeoff, which is a major advantage of the NALM.

In order to accurately decompose the aggregate load into its components, a model-based approach for describing individual appliances and their combination is used. These models suggest certain signatures which can be detected in the total load to indicate the activities of the separate components. This leads naturally to practical architectures and algorithms for the NALM. For full details, see the references below. We have implemented these ideas and carried out a number of initial field tests on residential loads to compare the NALM to traditional load monitoring techniques employed by electric utilities. Based on these tests, a commercial version of the NALM is being developed for widespread utility use.

Back to Table of Contents

Multistate Appliances

My recent research has focused on appliances which can be understood as multistate devices, using finite-state machine (FSM) models. There are three classes of appliance models from the NALM perspective:
ON/OFF (Two-state)
Appliances such as light bulbs or toasters, which are either on or off at any given moment. Early research focused on techniques for monitoring these.
Multistate
Appliances such as washing machines or dishwashers, with distinct types of ON states, e.g., fill, rinse, spin, pump, etc. Recent research has extended the methods to apply to the multi-state case.
Continuously variable
Appliances like light dimmers and variable-speed hand tools, with a continuous range of ON states. These are difficult to monitor nonintrusively, because they do not generate step changes in power.

To learn the FSM control structure of different multistate appliances, we have developed the portable instrumentation illustrated here. This is a new tool, which analyzes the behavior of an operating electrical load in a novel manner by automatically describing it with a finite-state model of its control structure. A personal-computer-based system collects samples of real and reactive power consumption over time, and automatically learns the control structure of the load in real time, drawing its finite-state diagram. The system also reports the load's state at each point in time, the total time spent in each state, and total energy consumption for each of the states.

One use of this tool is to provide a database of common appliance FSM structures for the NALM project. This research also has applications to behavioral analysis, energy monitoring, fault monitoring, fault analysis, and power quality analysis of many types of electrical loads, controllers, and power sources. Although only tested on residential loads and consumer appliances so far, the underlying methods should also work on commercial and industrial loads, e.g., HVAC control systems.

Three-way Lamp

As an example, if a three-way lamp is operated, a plot of power versus time shows plateaus at the low, medium, and high power levels. The patern-recognition algorithm in the instrument detects these, and constructs the FSM disaram shown, illustrating how the four states are cyclically connected.
 

Frost-free Refrigerator with Interior Lamp

A more complex example is this frost-free refrigerator. From the measured plots of real and reactive power, the six-state FSM shown is generated. The inner three states correspond to the light being off (the door closed) and the outer three occur when the light is on. The power plot shows how on/off cycles of the compressor are followed by a single defrost cycle in which the motor is off but a heater is on (so there is a large real power, but no reactive power). The FSM generated captures all this behavior
 
 

For details of the algorithm, discussion, and many more examples, see my paper ``Automatic Construction of Finite-State Load Behavior Models," listed in the references below.

Back to Table of Contents

NALM Development Status

The Electric Power Research Institute (EPRI) has sponsored NALM research since its conception in the early 1980's. EPRI has chosen Telog Instruments to commercialize the NALM into a research tool available to electric utilities. A beta-test program of the commercial version of the NALM is underway, and units are expected to be available to electric utilities in 1997. For exact availability information, contact: TelogSales@telog.com

Telog Instruments, Inc.
830 Canning Parkway
Victor, NY 14564-8940

(716) 742-3000

Back to Table of Contents

People to Contact

A number of people at a number of organizations are involved in research and development of NALM techniques.
George W. Hart
Originator and developer of the NALM. Originally at MIT, then at Columbia University, I was briefly at Hofstra University. Contact information and additional information regarding my research is available on my home page.
Lawrence Carmichael
Mr. Carmichael is the project supervisor at the Electric Power Research Institute (EPRI) in charge of the NALM. (415) 855-7982
Mark Malmendier
Mr. Malmendier is the product manager of Telog Instruments, Inc. in Rochester, New York. (716) 742-3000, email: TelogSales@telog.com
Leslie Norford and Steven Leeb
Profs. Norford and Leeb are engaged in research at MIT to explore the possibilities of extending NALM techniques to transient information in commercial buildings.
Jackie Lemmerhirt and Ralph Abbott
Plexus Research (in Acton, Massachusettes) is involved in coordinating the electric utility community to the NALM development efforts. Ms. Lemmerhirt (jlemmerhirt@plxs.com) is carrying out a project comparing NALM output with independent instrumentation in a number of test houses. This should provide solid data on the accuracy of the NALM. Mr. Abbott is president of Plexus.
Back to Table of Contents

References

A good introductory tutorial survey is the following article. (The above technical overview is excerpted from it.)

Two earlier technical reports are now available as PDFs:

A bibliography of early published papers concerning nonintrusive load monitoring is also available.