# A Scalable Real-Time Non-Intrusive Load Monitoring System for the Estimation of Household Appliance Power Consumption

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

- The proposed system is delay-free; once the appliance has been turned-on, the system can calculate its power in real-time.
- The proposed NILM algorithm is automatic, thus, no feedback is required by the user.

## 2. Proposed System

#### 2.1. Event Detection

_{d}. Down-sampling is applied for two main reasons: (a) the event detection algorithm becomes simpler, presenting less computational burden and (b) most of power changes are still easily identifiable assuming an 1 Hz sampling frequency. However, if two or more events occur almost simultaneously, e.g., in a period of less than a second, the algorithm detects these events as a single one. Considering that the probability of this scenario is very low, the frequency of 1 Hz has been selected. Next, the maximum power difference (MPD) for each second n is calculated as:

_{th}, which is determined in terms of the appliance rating power. This means that, at time instant n an event occurs if

_{tr}, is generated from P (100 × 6 = 600 samples). The pseudo-code for the process described is presented in Algorithm 1.

Algorithm 1: Event detection. |

#### 2.2. CNN Classifier

_{tr}, with a specific target appliance behavior, a CNN classifier is utilized. In this sense, for each target appliance, a dedicated CNN classifier is used, identifying P

_{tr}as positive when related to the target appliance or negative otherwise.

_{tr}by means of (3); the resulting normalized vector, P

_{norm}, is forwarded as input to the CNN model.

_{norm}. In particular, three consecutive 1-d convolutional layers are used in combination with an 1-d max-pooling layer. All convolutional layer parameters have been set to 32 filters, kernel size equal to 3, strides equal to 1, ’same’ padding, and rectified linear unit (ReLU) activation function. The ReLU function is defined as

**x**

_{conv}. Assuming that the size of

**x**

_{conv}is ${M}_{\mathrm{conv}}\times {N}_{\mathrm{conv}}$ and a single filter,

**f**, is of $3\times {N}_{\mathrm{conv}}$, the output of the convolution between

**x**

_{conv}and

**f**will be a ${M}_{\mathrm{conv}}\times 1$ matrix. The resulting

**y**

_{conv}is calculated as

**x**

_{conv}(0, n) and

**x**

_{conv}(M

_{conv}+1, n) are considered zero for any $n\in [1,\dots ,{N}_{\mathrm{conv}}]$ as a result of zero-padding. In our case, where 32 filters are used in a convolutional layer results

**y**

_{conv}, are stacked as columns, forming a ${M}_{\mathrm{conv}}\times 32$ matrix.

**x**

_{pool}, with size ${M}_{\mathrm{pool}}\times {N}_{\mathrm{pool}}$ is the max-pooling layer input matrix, the output,

**y**

_{pool}, has a size of $({M}_{\mathrm{pool}}/2)\times {N}_{\mathrm{pool}}$ and is calculated as

**w**, with size ${M}_{\mathrm{dense}}\times K$ and a bias vector,

**b**, with size K. Given an input vector,

**x**

_{dense}, with ${M}_{\mathrm{dense}}$ elements, the output

**y**

_{dense}of size K is calculated as

#### 2.3. Consumed Energy Estimation Algorithm

_{init}is considered equal to the appliance power consumption and assumed constant during the total time of operation of the appliance. When a power decrease between two consecutive seconds in P

_{d}inside the interval [0.8 P

_{init}, 1.2 P

_{init}] is detected, the appliance is considered to be turned-off. The pseudo-code of the energy consumption estimation algorithm is shown in Algorithm 2, having as input the time (in seconds), t, when the target appliance is turned-on and P

_{d}.

Algorithm 2: Energy consumption estimation. |

## 3. Evaluation Methodology

#### 3.1. Dataset

- Considering that the transient response starts at index s of z, a random number u in the interval [s − 500, s − 100] is selected, following uniform distribution. The selected sample is equal to z from index u to index u + 599.
- White Gaussian noise with mean value (μ) equal to 0 and standard deviation (σ) equal to 1 is added to the sample; 10 W maximum power is considered.

#### 3.2. Performance Metrics

#### 3.2.1. Metrics for Event Detection Evaluation

#### 3.2.2. Metrics for Classifier Evaluation

_{1}-score, defined in (10)–(13), respectively, are calculated

#### 3.2.3. Metrics for Overall NILM System Evaluation

_{1}-score to evaluate the predicted status of the appliance (ON or OFF). Thus, a sample (i.e., a time instant) is considered positive if the appliance is ON and negative if not. It should be mentioned that an appliance is considered turned-on if the measured active power is higher than 5 W. Additionally, for energy estimation, the mean absolute error (MAE) and the root mean square error (RMSE) in (14) and (15), respectively, are computed

## 4. Results

#### 4.1. Event Detection Evaluation

#### 4.2. Classification Evaluation

_{1}-score results are summarized in Table 4.

#### 4.3. Application on Residential Households

_{1}-score, MAE, RMSE and RE are calculated as well as their average considering the three households for 15 days. Results for the fridge, washing machine and microwave oven are shown in Table 5, Table 6 and Table 7, respectively. It can be generally observed that the proposed algorithm presents high accuracy regarding the power and energy estimates of the fridge and the microwave. On the contrary, the microwave oven recall metric is low. This can be attributed to the fact that the proposed methodology considers this appliance standby mode of operation as OFF. In fact, the power consumption during this period is low, thus, of trivial importance regarding energy consumption calculations. Regarding the washing machine results, the NILM system is designed to detect only the most energy-intensive process during the washing machine operation cycle, i.e., water heating mode of operation. For the rest of the operational cycles (non-detected), i.e., water pumping, drum spinning, rinsing, the appliance status is assumed OFF. The partial detection of the washing machine appliance is evident in Figure 6, resulting into low recall scores. Moreover, in the third household, the calculated low precision is due to the operation of appliances presenting similar transient response patterns, being misclassified as washing machine end-uses.

#### 4.4. Comparison with Other Methods

_{1}-score and accuracy calculations obtained by the proposed method are summarized in Table 8, Table 9 and Table 10 regarding the fridge, washing machine and microwave, respectively. The corresponding results (where available) reported in the relevant literature are also presented as well as the associated NILM technique, sampling frequency, and testing dataset. Note that, most of the literature state-of-the-art methods have been tested by using the well-known UK Domestic Appliance-Level Electricity (UK-DALE) [51] dataset. This dataset includes aggregated active power and appliance measurements of 0.167 Hz for several months, recorded for a small number of household installations. Moreover, the Reference Energy Disaggregation Data Set (REDD) [52] has been used in [21] to evaluate the LSTM algorithm performance; the sampling frequency is 1 Hz for mains and 0.333 Hz for the appliances. The proposed NILM system is tested by using an 100-Hz private dataset, since high-frequency sampling data are not provided in the above mentioned public datasets. It is important to stress out that in order to conduct a fair comparison between the different approaches, all metrics should be taken into consideration. However, this is not possible, since results for all metrics calculations are not always provided in the corresponding literature. Therefore, a direct comparison should be carried out with caution.

_{1}-score and accuracy, it can be realized that the proposed method is the second-best and first, respectively, among all examined solutions (where the corresponding metrics were available).

_{1}-score and accuracy metrics set the proposed method as the third- and fourth-best, respectively, among the examined solutions (where metrics were available).

_{1}-score and accuracy. Better results by other methods are observed only in terms of recall. This is due to the fact that the proposed system can not detect the microwave oven standby mode of operation. However, the power consumption during this period can be considered negligible. It is also important to note, that in NILM and from a user-experience point of view, precision is considered more important than recall; missing an appliance event is preferable than detecting an appliance event that has not actually occurred. In this sense, missing standby modes is more favored than predicting false microwave end-uses. The superiority of the proposed method for the analysis of the microwave oven is based on the following: (a) the microwave transient response pattern is unique, thus, it can be easily identified, and (b) the microwave oven end-use duration is short, varying from few seconds to minutes; thus, the number of the possible turning-off events caused from other appliances that may degrade the power estimation algorithm performance is very limited.

#### 4.5. Computational and Memory Efficiency

## 5. Discussion—Towards Scalable Real-Time NILM Services

- First of all, as expected, comes the accuracy metric. Accuracy usually refers to a weights-based combination of (i) correctly detected events, (ii) precise energy consumption estimation for the detected appliance events and (iii) minimized FP. Energy companies and electricity consumers usually trust a NILM service when its accuracy exceeds 90% and when they are not receiving reports for appliances/activities never actually occurred.
- Second comes the data resolution and as a result the data volumes required for an accurate NILM output. As mentioned above in Section 3, for real-time appliance identification sub-second data granularity is needed. Note that, most of the solutions presented in literature deal with kHz or even MHz of data. Considering as a rule of thumb that 1-s resolution data from separate phases in a 3-phase installation result in almost 1 GB of data being produced per year, we realize that moving into the kHz resolution areas makes data parsing, storing and analysing a rather complicated, costly and therefore non-scalable option.
- Next in the list comes the computational/RAM efficiency of such a service. Although the recent trend was to move everything to the cloud, now NILM vendors and energy companies realise that such a decision is not always the most cost-effective; the opposite actually. Running for example the whole service for ~100 k end users on the cloud can increase cloud operation costs that much, that there is no business case that can be built on top of a NILM layer, no matter how accurate that is. So, the key to unlock scalability opportunities here is to built a system that is so efficient that can run on the edge instead of the cloud.
- Strictly connected to the hardware constraints of the previous point comes the hardware cost. Traditionally sub-second data can be acquired only via a din meter hardware installed in the metering cabinet (it’s only recently that a few smart-meter manufacturers make >1 Hz resolutions available through their S1 port [53]). On the other hand, utilities and energy retail companies see NILM as a great customer engagement tool on top of which they can build value-added services and they usually tend to offer that as a freemium service. As a result, hardware cost has to be as low as possible and ideally within the companies customer retention and acquisition budgets.

## 6. Conclusions

## Author Contributions

## Funding

## Data Availability Statement

## Conflicts of Interest

## Abbreviations

AE | autoencoder |

BLUED | Building-Level fUlly-labeled dataset for Electricity Disaggregation |

C&I | commercial-industrial |

CNN | convolutional neural network |

CPU | central processing unit |

FN | false negative |

FNR | false negative rate |

FP | false positive |

FPR | false positive rate |

GRU | gated recurrent unit |

HMM | hidden Markov model |

ILM | intrusive load monitoring |

IoT | internet-of-things |

LSTM | long short-term memory |

MAE | mean absolute error |

MPD | maximum power difference |

NILM | non-intrusive load monitoring |

NLP | natural language processing |

PCNN | parallel CNN |

RE | relative error in total energy |

REDD | Reference Energy Disaggregation Data Set |

ReLU | rectified linear unit |

RMSE | root mean square error |

SAEDadd | self-attentive energy disaggregation with ‘dot’ attention mechanism |

SAEDdot | self-attentive energy disaggregation with ‘additive’ attention mechanism |

seq2point | sequence-to-point |

seq2seq | sequence-to-sequence |

TN | true negative |

TP | true positive |

TPR | true positive rate |

UK-DALE | UK Domestic Appliance-Level Electricity |

WGRU | window GRU |

## References

- Wang, Y.; Chen, Q.; Hong, T.; Kang, C. Review of Smart Meter Data Analytics: Applications, Methodologies, and Challenges. IEEE Trans. Smart Grid
**2019**, 10, 3125–3148. [Google Scholar] [CrossRef] [Green Version] - Hayes, B.P.; Prodanovic, M. State Forecasting and Operational Planning for Distribution Network Energy Management Systems. IEEE Trans. Smart Grid
**2016**, 7, 1002–1011. [Google Scholar] [CrossRef] - Huang, Y.; Wang, L.; Guo, W.; Kang, Q.; Wu, Q. Chance Constrained Optimization in a Home Energy Management System. IEEE Trans. Smart Grid
**2018**, 9, 252–260. [Google Scholar] [CrossRef] - Buzau, M.M.; Tejedor-Aguilera, J.; Cruz-Romero, P.; Gómez-Expósito, A. Detection of Non-Technical Losses Using Smart Meter Data and Supervised Learning. IEEE Trans. Smart Grid
**2019**, 10, 2661–2670. [Google Scholar] [CrossRef] - Chakraborty, S.; Das, S. Application of Smart Meters in High Impedance Fault Detection on Distribution Systems. IEEE Trans. Smart Grid
**2019**, 10, 3465–3473. [Google Scholar] [CrossRef] - Chatzigeorgiou, I.; Andreou, G. A systematic review on feedback research for residential energy behavior change through mobile and web interfaces. Renew. Sust. Energ. Rev.
**2021**, 135, 110187. [Google Scholar] [CrossRef] - Nalmpantis, C.; Vrakas, D. Machine learning approaches for non-intrusive load monitoring: From qualitative to quantitative comparation. Artif. Intel. Rev.
**2018**, 52. [Google Scholar] [CrossRef] - Ruano, A.; Hernandez, A.; Ureña, J.; Ruano, M.; Garcia, J. NILM Techniques for Intelligent Home Energy Management and Ambient Assisted Living: A Review. Energies
**2019**, 12, 2203. [Google Scholar] [CrossRef] [Green Version] - Hart, G.W. Non-intrusive appliance load monitoring. Proc. IEEE
**1992**, 80, 1870–1891. [Google Scholar] [CrossRef] - Kolter, J.Z.; Jaakkola, T. Approximate Inference in Additive Factorial HMMs with Application to Energy Disaggregation. In Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, La Palma, Canary Islands, 21–23 April 2012; Volume 22, pp. 1472–1482. Available online: http://proceedings.mlr.press/v22/zico12.html (accessed on 15 January 2021).
- Zoha, A.; Gluhak, A.; Nati, M.; Imran, M.A. Low-power appliance monitoring using Factorial Hidden Markov Models. In Proceedings of the 2013 IEEE Eighth International Conference on Intelligent Sensors, Sensor Networks and Information Processing, Melbourne, VIC, Australia, 2–5 April 2013; pp. 527–532. [Google Scholar] [CrossRef] [Green Version]
- Bonfigli, R.; Principi, E.; Fagiani, M.; Severini, M.; Squartini, S.; Piazza, F. Non-intrusive load monitoring by using active and reactive power in additive Factorial Hidden Markov Models. Appl. Energy
**2017**, 208, 1590–1607. [Google Scholar] [CrossRef] - Lu, T.; Xu, Z.; Huang, B. An Event-Based Nonintrusive Load Monitoring Approach: Using the Simplified Viterbi Algorithm. IEEE Pervasive Comput.
**2017**, 16, 54–61. [Google Scholar] [CrossRef] - Ji, T.Y.; Liu, L.; Wang, T.S.; Lin, W.B.; Li, M.S.; Wu, Q.H. Non-Intrusive Load Monitoring Using Additive Factorial Approximate Maximum a Posteriori Based on Iterative Fuzzy c-Means. IEEE Trans. Smart Grid
**2019**, 10, 6667–6677. [Google Scholar] [CrossRef] - Makonin, S.; Popowich, F.; Bajic, I.V.; Gill, B.; Bartram, L. Exploiting HMM Sparsity to Perform Online Real-Time Nonintrusive Load Monitoring. IEEE Trans. Smart Grid
**2016**, 7, 2575–2585. [Google Scholar] [CrossRef] - Kong, W.; Dong, Z.Y.; Hill, D.J.; Ma, J.; Zhao, J.H.; Luo, F.J. A Hierarchical Hidden Markov Model Framework for Home Appliance Modeling. IEEE Trans. Smart Grid
**2018**, 9, 3079–3090. [Google Scholar] [CrossRef] - Mueller, J.A.; Kimball, J.W. Accurate Energy Use Estimation for Nonintrusive Load Monitoring in Systems of Known Devices. IEEE Trans. Smart Grid
**2018**, 9, 2797–2808. [Google Scholar] [CrossRef] - Kim, J.; Le, T.T.H.; Kim, H. Non-intrusive Load Monitoring Based on Advanced Deep Learning and Novel Signature. Comput. Intel. Neurosc.
**2017**, 2017, 4216281. [Google Scholar] [CrossRef] - Kelly, J.; Knottenbelt, W. Neural NILM. In Proceedings of the 2nd ACM International Conference on Embedded Systems for Energy-Efficient Built Environments—BuildSys’15, Seoul, Korea, 4–5 November 2015. [Google Scholar] [CrossRef] [Green Version]
- Krystalakos, O.; Nalmpantis, C.; Vrakas, D. Sliding Window Approach for Online Energy Disaggregation Using Artificial Neural Networks. In Proceedings of the 10th Hellenic Conference on Artificial Intelligence, SETN ’18, Patra, Greece, 9–15 July 2018; Association for Computing Machinery: New York, NY, USA, 2018. [Google Scholar] [CrossRef]
- Mauch, L.; Yang, B. A new approach for supervised power disaggregation by using a deep recurrent LSTM network. In Proceedings of the 2015 IEEE Global Conference on Signal and Information Processing (GlobalSIP), Orlando, FL, USA, 14–16 December 2015; pp. 63–67. [Google Scholar] [CrossRef]
- Virtsionis Gkalinikis, N.; Nalmpantis, C.; Vrakas, D. Attention in Recurrent Neural Networks for Energy Disaggregation. In Discovery Science; Appice, A., Tsoumakas, G., Manolopoulos, Y., Matwin, S., Eds.; Springer International Publishing: Cham, Switzerland, 2020; pp. 551–565. [Google Scholar] [CrossRef]
- Kaselimi, M.; Protopapadakis, E.; Voulodimos, A.; Doulamis, N.; Doulamis, A. Multi-Channel Recurrent Convolutional Neural Networks for Energy Disaggregation. IEEE Access
**2019**, 7, 81047–81056. [Google Scholar] [CrossRef] - Sudoso, A.M.; Piccialli, V. Non-Intrusive Load Monitoring with an Attention-based Deep Neural Network. arXiv
**2019**, arXiv:1912.00759. [Google Scholar] - He, W.; Chai, Y. An Empirical Study on Energy Disaggregation via Deep Learning. In Proceedings of the 2016 2nd International Conference on Artificial Intelligence and Industrial Engineering (AIIE 2016), Beijing, China, 20–21 November 2016; Atlantis Press: Amsterdam, The Netherlands, 2016; pp. 338–342. [Google Scholar] [CrossRef] [Green Version]
- Zhang, C.; Zhong, M.; Wang, Z.; Goddard, N.; Sutton, C. Sequence-to-point learning with neural networks for non-intrusive load monitoring. arXiv
**2016**, arXiv:1612.09106. [Google Scholar] - Faustine, A.; Pereira, L.; Bousbiat, H.; Kulkarni, S. UNet-NILM: A Deep Neural Network for Multi-Tasks Appliances State Detection and Power Estimation in NILM. In Proceedings of the 5th International Workshop on Non-Intrusive Load Monitoring; NILM’20; Association for Computing Machinery: New York, NY, USA, 2020; pp. 84–88. [Google Scholar] [CrossRef]
- Devlin, M.A.; Hayes, B.P. Non-Intrusive Load Monitoring and Classification of Activities of Daily Living Using Residential Smart Meter Data. IEEE Trans. Consum. Electron.
**2019**, 65, 339–348. [Google Scholar] [CrossRef] - Penha, D.; Castro, A. Home Appliance Identification for NILM Systems Based on Deep Neural Networks. Int. J. Artif. Intell. Appl.
**2018**, 9, 69–80. [Google Scholar] [CrossRef] - Kong, W.; Dong, Z.Y.; Wang, B.; Zhao, J.; Huang, J. A Practical Solution for Non-Intrusive Type II Load Monitoring Based on Deep Learning and Post-Processing. IEEE Trans. Smart Grid
**2020**, 11, 148–160. [Google Scholar] [CrossRef] - Cui, G.; Liu, B.; Luan, W.; Yu, Y. Estimation of Target Appliance Electricity Consumption Using Background Filtering. IEEE Trans. Smart Grid
**2019**, 10, 5920–5929. [Google Scholar] [CrossRef] - Alcalá, J.; Ureña, J.; Hernández, Á.; Gualda, D. Event-Based Energy Disaggregation Algorithm for Activity Monitoring From a Single-Point Sensor. IEEE Trans. Instrum. Meas.
**2017**, 66, 2615–2626. [Google Scholar] [CrossRef] - Faustine, A.; Pereira, L. Improved Appliance Classification in Non-Intrusive Load Monitoring Using Weighted Recurrence Graph and Convolutional Neural Networks. Energies
**2020**, 13, 3374. [Google Scholar] [CrossRef] - Yang, D.; Gao, X.; Kong, L.; Pang, Y.; Zhou, B. An Event-Driven Convolutional Neural Architecture for Non-Intrusive Load Monitoring of Residential Appliance. IEEE Trans. Consum. Electron.
**2020**, 66, 173–182. [Google Scholar] [CrossRef] - De Baets, L.; Ruyssinck, J.; Develder, C.; Dhaene, T.; Deschrijver, D. Appliance classification using VI trajectories and convolutional neural networks. Energ. Build.
**2018**, 158, 32–36. [Google Scholar] [CrossRef] [Green Version] - Liu, Y.; Wang, X.; You, W. Non-Intrusive Load Monitoring by Voltage—Current Trajectory Enabled Transfer Learning. IEEE Trans. Smart Grid
**2019**, 10, 5609–5619. [Google Scholar] [CrossRef] - Hassan, T.; Javed, F.; Arshad, N. An Empirical Investigation of V-I Trajectory Based Load Signatures for Non-Intrusive Load Monitoring. IEEE Trans. Smart Grid
**2014**, 5, 870–878. [Google Scholar] [CrossRef] [Green Version] - Du, L.; He, D.; Harley, R.G.; Habetler, T.G. Electric Load Classification by Binary Voltage—Current Trajectory Mapping. IEEE Trans. Smart Grid
**2016**, 7, 358–365. [Google Scholar] [CrossRef] - Faustine, A.; Pereira, L. Multi-Label Learning for Appliance Recognition in NILM Using Fryze-Current Decomposition and Convolutional Neural Network. Energies
**2020**, 13, 4154. [Google Scholar] [CrossRef] - Krizhevsky, A.; Sutskever, I.; Hinton, G. ImageNet Classification with Deep Convolutional Neural Networks. Neural Inf. Process Syst.
**2012**, 25. [Google Scholar] [CrossRef] - Chang, H.H.; Lin, L.S.; Chen, N.; Lee, W.J. Particle Swarm Optimization based non-intrusive demand monitoring and load identification in smart meters. In Proceedings of the 2012 IEEE Industry Applications Society Annual Meeting, Las Vegas, NV, USA, 7–11 October 2012; pp. 1–8. [Google Scholar] [CrossRef]
- Chang, H.H.; Chen, K.L.; Tsai, Y.P.; Lee, W.J. A New Measurement Method for Power Signatures of Nonintrusive Demand Monitoring and Load Identification. IEEE Trans. Ind. Appl.
**2012**, 48, 764–771. [Google Scholar] [CrossRef] - Zoha, A.; Gluhak, A.; Imran, M.; Rajasegarar, S. Non-Intrusive Load Monitoring Approaches for Disaggregated Energy Sensing: A Survey. Sensors
**2012**, 12, 16838–16866. [Google Scholar] [CrossRef] [Green Version] - Revuelta Herrero, J.; Lozano Murciego, A.; Barriuso, A.; Hernández de la Iglesia, D.; Villarrubia, G.; Corchado Rodríguez, J.; Carreira, R. Non Intrusive Load Monitoring (NILM): A State of the Art. In Trends in Cyber-Physical Multi-Agent Systems; Springer International Publishing: Cham, Switzerland, 2018; pp. 125–138. [Google Scholar] [CrossRef]
- LeCun, Y.; Bengio, Y.; Hinton, G. Deep Learning. Nature
**2015**, 521, 436–444. [Google Scholar] [CrossRef] - Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J. Mach. Learn. Res.
**2014**, 15, 1929–1958. Available online: http://jmlr.org/papers/v15/srivastava14a.html (accessed on 15 January 2021). - Kingma, D.P.; Ba, J. Adam: A Method for Stochastic Optimization. arXiv
**2017**, arXiv:1412.6980. [Google Scholar] - Anderson, K.; Ocneanu, A.; Benitez, D.; Carlson, D.; Rowe, A.; Bergés, M. BLUED: A Fully Labeled Public Dataset for Event-Based Non-Intrusive Load Monitoring Research. In Proceedings of the 2nd KDD Workshop on Data Mining Applications in Sustainability (SustKDD), Beijing, China, 12 August 2012. [Google Scholar]
- Lu, M.; Li, Z. A Hybrid Event Detection Approach for Non-Intrusive Load Monitoring. IEEE Trans. Smart Grid
**2020**, 11, 528–540. [Google Scholar] [CrossRef] [Green Version] - Bergés, M.; Goldman, E.; Matthews, H.; Soibelman, L.; Anderson, K. User-Centered Nonintrusive Electricity Load Monitoring for Residential Buildings. J. Comput. Civ. Eng.
**2011**, 25, 471–480. [Google Scholar] [CrossRef] - Kelly, J.; Knottenbelt, W. The UK-DALE dataset, domestic appliance-level electricity demand and whole-house demand from five UK homes. Sci. Data
**2015**, 2. [Google Scholar] [CrossRef] [Green Version] - Kolter, J.; Johnson, M. REDD: A Public Data Set for Energy Disaggregation Research. In Proceedings of the SustKDD Workshop on Data Mining Applications in Sustainability, San Diego, CA, USA, 21 August 2011. [Google Scholar]
- Fluvius. User Ports of Smart Meters Explanatory Note on the Possibilities for Product Developers; Technical Report; Melle, Belgium, 2019; Available online: https://www.fluvius.be/sites/fluvius/files/2019-07/1901-fluvius-technical-specification-user-ports-digital-meter.pdf (accessed on 15 January 2021).

**Figure 3.**Comparison of the turn-on transient response with sampling frequency at 100 Hz and 1 Hz for (

**a**) fridge, (

**b**) washing machine, (

**c**) microwave oven, (

**d**) stove and (

**e**) heat pump dryer.

**Figure 6.**Power estimation for the selected appliances in real households. Time-series of (

**a**) aggregated power, (

**b**) actual target appliance power, (

**c**) estimated power for fridge; (

**d**) aggregated power, (

**e**) actual target appliance power, (

**f**) estimated power for washing machine; (

**g**) aggregated power, (

**h**) actual target appliance power, (

**i**) estimated power for microwave.

Appliance | Number of Transient Responses |
---|---|

Fridge | 132 |

Dishwasher | 171 |

Heat pump | 202 |

Washing machine | 135 |

Oven | 82 |

Stove | 148 |

Heat pump dryer (drum spinning) | 54 |

Heat pump dryer (heating) | 42 |

Microwave | 290 |

Appliance | Training | Validation | Testing |
---|---|---|---|

Fridge | 2400 | 780 | 780 |

Washing machine | 2430 | 810 | 810 |

Microwave | 5220 | 1740 | 1740 |

Reference | TPR | FPR | FNR |
---|---|---|---|

Proposed | 94.400% | 0.003% | 5.600% |

[32] | 94.000% | 0.088% | 6.000% |

[49] | 96.700% | 0.810% | 3.300% |

[50] | 94.130% | 0.260% | 5.870% |

Appliance | Accuracy | Precision | Recall | F_{1}-Score |
---|---|---|---|---|

Fridge | 0.978 | 0.984 | 0.972 | 0.978 |

Washing machine | 0.872 | 0.875 | 0.867 | 0.871 |

Microwave | 0.992 | 0.986 | 0.999 | 0.992 |

House | Accuracy | Precision | Recall | F_{1}-Score | MAE (W) | RMSE (W) | E (kWh) | $\hat{\mathit{E}}$ (kWh) | RE |
---|---|---|---|---|---|---|---|---|---|

1 | 0.86 | 0.94 | 0.77 | 0.85 | 12.97 | 38.67 | 14.26 | 11.74 | 0.18 |

2 | 0.89 | 0.96 | 0.74 | 0.84 | 7.05 | 17.14 | 6.19 | 4.24 | 0.32 |

3 | 0.98 | 1.00 | 0.90 | 0.95 | 3.68 | 12.89 | 4.74 | 5.05 | 0.06 |

Average | 0.91 | 0.97 | 0.80 | 0.88 | 7.90 | 22.90 | - | - | 0.19 |

House | Accuracy | Precision | Recall | F_{1}-Score | MAE (W) | RMSE (W) | E (kWh) | $\hat{\mathit{E}}$ (kWh) | RE |
---|---|---|---|---|---|---|---|---|---|

1 | 0.94 | 0.99 | 0.24 | 0.39 | 20.86 | 168.96 | 19.85 | 12.83 | 0.35 |

2 | 0.97 | 0.94 | 0.24 | 0.38 | 10.02 | 116.83 | 10.44 | 8.44 | 0.19 |

3 | 0.93 | 0.42 | 0.13 | 0.20 | 24.27 | 183.42 | 7.27 | 12.61 | 0.42 |

Average | 0.95 | 0.78 | 0.20 | 0.32 | 18.38 | 156.40 | - | - | 0.32 |

House | Accuracy | Precision | Recall | F_{1}-Score | MAE (W) | RMSE (W) | E (kWh) | $\hat{\mathit{E}}$ (kWh) | RE |
---|---|---|---|---|---|---|---|---|---|

1 | 1.00 | 0.94 | 0.52 | 0.67 | 1.28 | 36.55 | 2.16 | 2.13 | 0.01 |

2 | 1.00 | 0.99 | 0.46 | 0.63 | 1.08 | 37.63 | 2.10 | 1.90 | 0.10 |

3 | 0.99 | 0.82 | 0.47 | 0.60 | 2.61 | 57.27 | 2.19 | 2.57 | 0.15 |

Average | 1.00 | 0.92 | 0.48 | 0.63 | 1.66 | 43.82 | - | - | 0.09 |

**Table 8.**Comparison results among existing non-intrusive load monitoring (NILM) solutions for fridge identification and energy consumption estimation.

Reference | Method | Sampling Frequency | Dataset | MAE | RE | Precision | Recall | F_{1}-Score | Accuracy |
---|---|---|---|---|---|---|---|---|---|

Proposed | 100 Hz | private | 7.90 | 0.19 | 0.97 | 0.80 | 0.88 | 0.91 | |

[19] | Autoencoder | 0.167 Hz | UK-DALE | 26.00 | 0.38 | 0.85 | 0.88 | 0.87 | 0.90 |

[19] | CNN | 0.167 Hz | UK-DALE | 18.00 | 0.13 | 0.79 | 0.86 | 0.82 | 0.87 |

[19] | LSTM | 0.167 Hz | UK-DALE | 36.00 | 0.25 | 0.72 | 0.77 | 0.74 | 0.81 |

[20] | LSTM | 0.167 Hz | UK-DALE | 51.00 | 0.21 | 0.45 | 0.51 | 0.47 | 0.60 |

[20] | GRU | 0.167 Hz | UK-DALE | 51.00 | 0.26 | 0.46 | 0.75 | 0.57 | 0.60 |

[20] | seq2point | 0.167 Hz | UK-DALE | 51.00 | 0.29 | 0.42 | 0.74 | 0.53 | 0.54 |

[21] | LSTM | 0.333 Hz | REDD | - | - | 0.91 | 0.96 | 0.93 | - |

[22] | WGRU | 0.167 Hz | UK-DALE | 28.46 | 0.13 | - | - | 0.82 | - |

[22] | SAEDdot | 0.167 Hz | UK-DALE | 35.25 | 0.60 | - | - | 0.62 | - |

[22] | SAEDadd | 0.167 Hz | UK-DALE | 32.31 | 0.65 | - | - | 0.66 | - |

[25] | PCNN AE | 0.167 Hz | UK-DALE | 3.46 | - | - | - | - | - |

[25] | PCNN LSTM | 0.167 Hz | UK-DALE | 3.22 | - | - | - | - | - |

[26] | seq2seq | 0.167 Hz | UK-DALE | 24.49 | - | - | - | - | - |

[26] | seq2point | 0.167 Hz | UK-DALE | 20.89 | - | - | - | - | - |

[27] | UNet | 0.167 Hz | UK-DALE | 15.12 | - | - | - | - | - |

**Table 9.**Comparison results among existing NILM solutions for washing machine identification and energy consumption estimation.

Reference | Method | Sampling Frequency | Dataset | MAE | RE | Precision | Recall | F_{1}-score | Accuracy |
---|---|---|---|---|---|---|---|---|---|

Proposed | 100 Hz | private | 18.38 | 0.32 | 0.78 | 0.20 | 0.32 | 0.95 | |

[19] | Autoencoder | 0.167 Hz | UK-DALE | 24.00 | 0.48 | 0.07 | 1.00 | 0.13 | 0.82 |

[19] | CNN | 0.167 Hz | UK-DALE | 11.00 | 0.74 | 0.29 | 0.24 | 0.27 | 0.98 |

[19] | LSTM | 0.167 Hz | UK-DALE | 109.00 | 0.91 | 0.01 | 0.73 | 0.03 | 0.23 |

[20] | LSTM | 0.167 Hz | UK-DALE | 25.00 | 0.35 | 0.16 | 0.56 | 0.24 | 0.95 |

[20] | GRU | 0.167 Hz | UK-DALE | 30.00 | 0.58 | 0.22 | 0.54 | 0.31 | 0.96 |

[20] | seq2point | 0.167 Hz | UK-DALE | 17.00 | 0.28 | 0.26 | 0.55 | 0.35 | 0.97 |

[22] | WGRU | 0.167 Hz | UK-DALE | 10.45 | 0.43 | - | - | 0.34 | - |

[22] | SAEDdot | 0.167 Hz | UK-DALE | 13.10 | 0.34 | - | - | 0.30 | - |

[22] | SAEDadd | 0.167 Hz | UK-DALE | 22.01 | 0.53 | - | - | 0.30 | - |

[25] | PCNN AE | 0.167 Hz | UK-DALE | 83.40 | - | - | - | - | - |

[25] | PCNN LSTM | 0.167 Hz | UK-DALE | 73.16 | - | - | - | - | - |

[26] | seq2seq | 0.167 Hz | UK-DALE | 10.15 | - | - | - | - | - |

[26] | seq2point | 0.167 Hz | UK-DALE | 12.66 | - | - | - | - | - |

[27] | UNet | 0.167 Hz | UK-DALE | 11.51 | - | - | - | - | - |

**Table 10.**Comparison results among existing NILM solutions for microwave identification and energy consumption estimation.

Reference | Method | Sampling Frequency | Dataset | MAE | RE | Precision | Recall | F_{1}-Score | Accuracy |
---|---|---|---|---|---|---|---|---|---|

Proposed | 100 Hz | private | 1.66 | 0.09 | 0.92 | 0.48 | 0.63 | 1.00 | |

[19] | Autoencoder | 0.167 Hz | UK-DALE | 9.00 | 0.73 | 0.15 | 0.94 | 0.26 | 0.99 |

[19] | CNN | 0.167 Hz | UK-DALE | 6.00 | 0.50 | 0.14 | 0.40 | 0.21 | 0.99 |

[19] | LSTM | 0.167 Hz | UK-DALE | 20.00 | 0.88 | 0.07 | 0.99 | 0.13 | 0.98 |

[20] | LSTM | 0.167 Hz | UK-DALE | 86.00 | 0.10 | 0.01 | 0.45 | 0.02 | 0.93 |

[20] | GRU | 0.167 Hz | UK-DALE | 97.00 | 0.07 | 0.02 | 0.75 | 0.04 | 0.93 |

[20] | seq2point | 0.167 Hz | UK-DALE | 103.00 | 0.16 | 0.01 | 0.79 | 0.03 | 0.91 |

[21] | LSTM | 0.333 Hz | REDD | - | - | 0.50 | 0.05 | 0.09 | - |

[22] | WGRU | 0.167 Hz | UK-DALE | 4.36 | 0.25 | - | - | 0.44 | - |

[22] | SAEDdot | 0.167 Hz | UK-DALE | 5.97 | 0.19 | - | - | 0.25 | - |

[22] | SAEDadd | 0.167 Hz | UK-DALE | 5.98 | 0.17 | - | - | 0.26 | - |

[25] | PCNN AE | 0.167 Hz | UK-DALE | 27.50 | - | - | - | - | - |

[25] | PCNN LSTM | 0.167 Hz | UK-DALE | 9.42 | - | - | - | - | - |

[26] | seq2seq | 0.167 Hz | UK-DALE | 13.62 | - | - | - | - | - |

[26] | seq2point | 0.167 Hz | UK-DALE | 8.67 | - | - | - | - | - |

[27] | UNet | 0.167 Hz | UK-DALE | 6.48 | - | - | - | - | - |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Athanasiadis, C.; Doukas, D.; Papadopoulos, T.; Chrysopoulos, A.
A Scalable Real-Time Non-Intrusive Load Monitoring System for the Estimation of Household Appliance Power Consumption. *Energies* **2021**, *14*, 767.
https://doi.org/10.3390/en14030767

**AMA Style**

Athanasiadis C, Doukas D, Papadopoulos T, Chrysopoulos A.
A Scalable Real-Time Non-Intrusive Load Monitoring System for the Estimation of Household Appliance Power Consumption. *Energies*. 2021; 14(3):767.
https://doi.org/10.3390/en14030767

**Chicago/Turabian Style**

Athanasiadis, Christos, Dimitrios Doukas, Theofilos Papadopoulos, and Antonios Chrysopoulos.
2021. "A Scalable Real-Time Non-Intrusive Load Monitoring System for the Estimation of Household Appliance Power Consumption" *Energies* 14, no. 3: 767.
https://doi.org/10.3390/en14030767