pith. sign in

arxiv: 1907.06299 · v2 · pith:YRKWBC57new · submitted 2019-07-15 · 📡 eess.SP

Universal Non-Intrusive Load Monitoring (UNILM) Using Filter Pipelines, Probabilistic Knapsack, and Labelled Partition Maps

Alejandro Rodriguez-Silva , Stephen Makonin This is my paper

Pith reviewed 2026-05-24 21:40 UTC · model grok-4.3

classification 📡 eess.SP
keywords NILMnon-intrusive load monitoringenergy disaggregationprobabilistic knapsackfilter pipelinespartition mapsunsupervised learningsmart meter data
0
0 comments X

The pith

A universal NILM method using filter pipelines, probabilistic knapsack, and labelled partition maps tracks appliances unsupervised and accounts for 93.7% of aggregate energy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents a method for non-intrusive load monitoring that functions unsupervised and adapts to different appliance configurations in various countries. Data is preprocessed with a filter pipeline, appliances are modeled as Gaussian distributions and disaggregated using a probabilistic knapsack algorithm, and labelled partition maps identify the appliances and quantify their energy consumption. The approach is tested on aggregate smart meter signals. Results indicate that relatively complex appliance signals can be tracked while accounting for 93.7% of the total energy consumed. Readers would care if this enables widespread energy savings and environmental benefits without the need for individual appliance sensors or region-specific data.

Core claim

The paper establishes that the UNILM method, consisting of an advanced filter pipeline to preprocess the data, a Gaussian appliance model with a probabilistic knapsack algorithm to disaggregate the aggregate smart meter signal, and partition maps to label which appliances were found and how much energy they use, can operate without prior knowledge and independent of regional differences in appliance mixes.

What carries the argument

Labelled partition maps that enable appliance identification and energy quantification independent of country or region.

If this is right

  • The system discovers appliances without prior knowledge or training data.
  • It disaggregates signals to track complex appliance usage.
  • The method accounts for 93.7% of total aggregate energy consumed.
  • It runs independently across different countries and regions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This approach might be extended to real-time energy management in smart homes.
  • Testing on additional international datasets could validate broader applicability.
  • Combining with other monitoring techniques could improve accuracy for rare appliances.

Load-bearing premise

The Gaussian appliance model and labelled partition maps generalize to different appliance mixes and operational characteristics in various countries without prior knowledge or training data specific to those regions.

What would settle it

Running the method on aggregate power data from a different country with unseen appliance types and checking if energy tracking remains near 93.7% without any retraining.

Figures

Figures reproduced from arXiv: 1907.06299 by Alejandro Rodriguez-Silva, Stephen Makonin.

Figure 1
Figure 1. Figure 1: Sensing appliance power usage and consumption without sensors. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Results (left) of a two-year case study using AMPds [5] where a 1980’s wireless garage door lift was found to have consumed a large amount of [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparing the raw 1Hz aggregate signal (left) to output of our filter pipeline (right). As expected, clean transients and flat steady-states. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: An example of a coloured partition map to auto-label North American appliances. Each country/region would have there own specific partition map [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparing the ground truth sub-metered appliance signals (top) [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
read the original abstract

Being able to track appliances energy usage without the need of sensors can help occupants reduce their energy consumption to help save the environment all while saving money. Non-intrusive load monitoring (NILM) tries to do just that. One of the hardest problems NILM faces is the ability to run unsupervised -- discovering appliances without prior knowledge -- and to run independent of the differences in appliance mixes and operational characteristics found in various countries and regions. We propose a solution that can do this with the use of an advanced filter pipeline to preprocess the data, a Gaussian appliance model with a probabilistic knapsack algorithm to disaggregate the aggregate smart meter signal, and partition maps to label which appliances were found and how much energy they use no matter the country/region. Experimental results show that relatively complex appliance signals can be tracked accounting for 93.7% of the total aggregate energy consumed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes Universal Non-Intrusive Load Monitoring (UNILM), a method for appliance energy disaggregation that operates without prior knowledge or region-specific training data. It combines an advanced filter pipeline for preprocessing, a Gaussian appliance model with a probabilistic knapsack algorithm for disaggregating the aggregate signal, and labelled partition maps to identify appliances and account for their energy use. The central claim is that this approach tracks relatively complex appliance signals while accounting for 93.7% of total aggregate energy consumed across different countries and regions.

Significance. If the 93.7% result and the claimed generalization hold under proper multi-region validation, the work would be significant for NILM by offering an unsupervised framework that avoids the need for supervised retraining when appliance mixes or usage patterns change across regions.

major comments (2)
  1. [Abstract] Abstract: the central claim that the method accounts for 93.7% of aggregate energy is presented without any description of the dataset(s), number of appliances, experimental protocol, number of trials, or baseline comparisons, rendering it impossible to assess whether the Gaussian model and partition maps actually support the no-training-data universality claim.
  2. [Abstract] Abstract / experimental results: no evidence is provided that validation used aggregate signals drawn from multiple countries with differing appliance mixes, voltage standards, or usage patterns; the reported figure therefore does not test the distributional-shift robustness that the universality claim requires.
minor comments (1)
  1. [Abstract] The abstract contains informal phrasing (e.g., 'tries to do just that') that is atypical for a technical journal submission.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for your review and the recommendation for major revision. We address the major comments on the abstract below and will make the necessary revisions to improve clarity regarding the experimental setup and the scope of the universality claim.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the method accounts for 93.7% of aggregate energy is presented without any description of the dataset(s), number of appliances, experimental protocol, number of trials, or baseline comparisons, rendering it impossible to assess whether the Gaussian model and partition maps actually support the no-training-data universality claim.

    Authors: We agree that the abstract is too concise and omits key experimental details. In the revised version we will expand the abstract to briefly describe the public NILM datasets used, the typical number of appliances per household, the unsupervised experimental protocol (no region-specific training data), the number of trials, and note the absence of direct baseline comparisons given the focus on the unsupervised setting. revision: yes

  2. Referee: [Abstract] Abstract / experimental results: no evidence is provided that validation used aggregate signals drawn from multiple countries with differing appliance mixes, voltage standards, or usage patterns; the reported figure therefore does not test the distributional-shift robustness that the universality claim requires.

    Authors: The reported experiments use publicly available NILM datasets that contain households with varying appliance mixes and usage patterns. However, these datasets are drawn primarily from one region and therefore do not explicitly test robustness across countries with different voltage standards. The universality claim rests on the method's design (no training data required), not on the breadth of the current empirical evaluation. We will revise the manuscript to clarify this distinction and add a limitations discussion on multi-region validation. revision: partial

Circularity Check

0 steps flagged

No derivation chain present; experimental claim is self-contained

full rationale

The paper presents a method for universal NILM via filter pipelines, Gaussian appliance model, probabilistic knapsack, and labelled partition maps, with the sole quantitative claim being an experimental result of 93.7% aggregate energy accounted for. No equations, derivations, or mathematical steps appear in the abstract or described content. The generalization claim is framed as an empirical outcome rather than a first-principles derivation that reduces to fitted inputs or self-citations by construction. Absent any load-bearing mathematical chain that collapses to its own definitions or parameters, the result does not exhibit circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient detail to identify or populate free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5685 in / 1045 out tokens · 27488 ms · 2026-05-24T21:40:58.174202+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    Nonintrusive appliance load monitoring,

    G. Hart, “Nonintrusive appliance load monitoring,” Proceedings of the IEEE, vol. 80, no. 12, pp. 1870–1891, Dec 1992

    work page 1992

  2. [2]

    Impact of energy disaggregation on consumer behavior,

    P. Chakravarty and A. Gupta, “Impact of energy disaggregation on consumer behavior,” in UC Berkeley: Behavior, Energy and Climate Change Conference, 2013

    work page 2013

  3. [3]

    Advanced me- tering initiatives and residential feedback programs: a meta-review for household electricity-saving opportunities,

    K. Ehrhardt-Martinez, K. Donnelly, S. Laitner et al. , “Advanced me- tering initiatives and residential feedback programs: a meta-review for household electricity-saving opportunities,” in Report Number E105 . ACEEE Washington, DC, 2010

    work page 2010

  4. [4]

    Is disaggregation the holy grail of energy efficiency? the case of electricity,

    K. Armel, A. Gupta, G. Shrimali, and A. Albert, “Is disaggregation the holy grail of energy efficiency? the case of electricity,” Energy Policy, vol. 52, pp. 213–234, 2013

    work page 2013

  5. [5]

    Electricity, water, and natural gas consumption of a residential house in Canada from 2012 to 2014,

    S. Makonin, B. Ellert, I. Baji ´c, and F. Popowich, “Electricity, water, and natural gas consumption of a residential house in Canada from 2012 to 2014,” Scientific Data, vol. 3, no. 160037, pp. 1–12, 06 2016

    work page 2012

  6. [6]

    REDD: A public data set for energy disaggregation research,

    J. Kolter and M. Johnson, “REDD: A public data set for energy disaggregation research,” in Workshop on Data Mining Applications in Sustainability (SIGKDD), San Diego, CA , vol. 25, 2011, pp. 59–62

    work page 2011

  7. [7]

    An electrical load mea- surements dataset of United Kingdom households from a two-year longitudinal study,

    D. Murray, L. Stankovic, and V . Stankovic, “An electrical load mea- surements dataset of United Kingdom households from a two-year longitudinal study,” Scientific Data, vol. 4, pp. 160 122 EP –, 01 2017

    work page 2017

  8. [8]

    RAE: The Rainforest Automa- tion energy dataset for smart grid meter data analysis,

    S. Makonin, Z. Wang, and C. Tumpach, “RAE: The Rainforest Automa- tion energy dataset for smart grid meter data analysis,”data, vol. 3, no. 1, p. 8, 2018

    work page 2018

  9. [9]

    Ex- ploiting hmm sparsity to perform online real-time nonintrusive load monitoring,

    S. Makonin, F. Popowich, I. V . Baji ´c, B. Gill, and L. Bartram, “Ex- ploiting hmm sparsity to perform online real-time nonintrusive load monitoring,” IEEE Transactions on Smart Grid, vol. 7, no. 6, pp. 2575– 2585, 2016

    work page 2016

  10. [10]

    Neural variational identification and filtering for stochastic non-linear dynamical systems with applica- tion to non-intrusive load monitoring,

    H. Lange, M. Berg ´es, and Z. Kolter, “Neural variational identification and filtering for stochastic non-linear dynamical systems with applica- tion to non-intrusive load monitoring,” in ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 2019, pp. 8340–8344

    work page 2019

  11. [11]

    Multi label restricted boltz- mann machine for non-intrusive load monitoring,

    S. Verma, S. Singh, and A. Majumdar, “Multi label restricted boltz- mann machine for non-intrusive load monitoring,” in ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 2019, pp. 8345–8349

    work page 2019

  12. [12]

    wavenilm: A causal neural network for power disaggregation from the complex power signal,

    A. Harell, S. Makonin, and I. V . Baji ´c, “wavenilm: A causal neural network for power disaggregation from the complex power signal,” in ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    work page 2019

  13. [13]

    On the applicability of correlation filters for appliance detection in smart meter readings,

    C. Klemenjak and W. Elmenreich, “On the applicability of correlation filters for appliance detection in smart meter readings,” in 2017 IEEE International Conference on Smart Grid Communications (SmartGrid- Comm), Oct 2017, pp. 171–176

    work page 2017

  14. [14]

    Improving event- based non-intrusive load monitoring using graph signal processing,

    B. Zhao, K. He, L. Stankovic, and V . Stankovic, “Improving event- based non-intrusive load monitoring using graph signal processing,” IEEE Access, vol. 6, pp. 53 944–53 959, 2018

    work page 2018

  15. [15]

    An unsupervised training method for non-intrusive appliance load monitoring,

    O. Parson, S. Ghosh, M. Weal, and A. Rogers, “An unsupervised training method for non-intrusive appliance load monitoring,” Artificial Intelligence, vol. 217, pp. 1–19, 2014

    work page 2014

  16. [16]

    Transferability of neural network approaches for low-rate energy dis- aggregation,

    D. Murray, L. Stankovic, V . Stankovic, S. Lulic, and S. Sladojevic, “Transferability of neural network approaches for low-rate energy dis- aggregation,” in ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , May 2019, pp. 8330–8334

    work page 2019

  17. [17]

    Evolving non-intrusive load monitoring,

    D. Egarter, A. Sobe, and W. Elmenreich, “Evolving non-intrusive load monitoring,” in European Conference on the Applications of Evolution- ary Computation. Springer, 2013, pp. 182–191

    work page 2013

  18. [18]

    Leveraging smart meter data to recognize home appliances,

    M. Weiss, A. Helfenstein, F. Mattern, and T. Staake, “Leveraging smart meter data to recognize home appliances,” in 2012 IEEE International Conference on Pervasive Computing and Communications. IEEE, 2012, pp. 190–197

    work page 2012

  19. [19]

    Bilateral filtering for gray and color images

    C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images.” in Iccv, vol. 98, no. 1, 1998, p. 2

    work page 1998

  20. [20]

    Scale-space and edge detection using anisotropic diffusion,

    P. Perona and J. Malik, “Scale-space and edge detection using anisotropic diffusion,” IEEE Transactions on pattern analysis and ma- chine intelligence, vol. 12, no. 7, pp. 629–639, 1990

    work page 1990

  21. [21]

    Domain transform for edge-aware image and video processing,

    E. S. Gastal and M. M. Oliveira, “Domain transform for edge-aware image and video processing,” in ACM Transactions on Graphics (ToG), vol. 30, no. 4. ACM, 2011, p. 69

    work page 2011

  22. [22]

    Knapsack problems. 2004,

    H. Kellerer, U. Pferschy, and D. Pisinger, “Knapsack problems. 2004,” 2003

    work page 2004

  23. [23]

    Approximate inference in additive factorial hmms with application to energy disaggregation,

    J. Z. Kolter and T. Jaakkola, “Approximate inference in additive factorial hmms with application to energy disaggregation,” in Artificial Intelli- gence and Statistics , 2012, pp. 1472–1482

    work page 2012

  24. [24]

    Nonintrusive load monitoring (NILM) performance evaluation,

    S. Makonin and F. Popowich, “Nonintrusive load monitoring (NILM) performance evaluation,” Energy Efficiency, vol. 8, no. 4, pp. 809–814, Jul 2015

    work page 2015