A Benchmark Framework to Evaluate Energy Disaggregation Solutions

Symeonidis, Nikolaos; Nalmpantis, Christoforos; Vrakas, Dimitris

doi:10.1007/978-3-030-20257-6_2

Cited by 7 publications

(20 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For a fair comparison the best environmental setup is found for each of the four models that are compared. Then, utilizing the benchmark framework of Symeonidis et al [5], it is demonstrated that the proposed model achieves close to state-of-the-art results, whereas it remains computational efficient, it has less learning parameters and requires relatively small storage space.…”

Section: Introductionmentioning

confidence: 93%

“…The final experiments take the window size into account and all the models are adjusted finding their best performing window per appliance. The evaluation is based on the benchmark framework that is proposed by Symeonidis et al [5], showcasing that the proposed architecture performs on par with other strong baselines, whereas it has less learning parameters, faster inference and training time and reduced size.…”

Section: Related Workmentioning

confidence: 99%

“…The main advantage of Fourier transform is lower computational complexity, faster inference speed and smaller size of a trained model. Regarding the evaluation and comparison of the proposed model against existing ones, the process aligns with the benchmark framework that is proposed by Symeonidis et al [5] and includes four basic scenarios. In the first case the models are trained and tested on the same house at different time periods.…”

Section: On House 1 From Uk-dalementioning

confidence: 99%

“…The complexity of the environmental setup makes the reproducibility of NILM experiments a hard task. In order to overcome the comparability issue, Symeonidis et al [5] propose a benchmark framework that describes different scenarios of testing NILM algorithms. Batra et al [6] try to tackle reproducibility issues by providing the implementation of nine different disaggregation algorithms along with state-of-the art experimental results.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Neural Fourier Energy Disaggregation

Nalmpantis

Gkalinikis

Vrakas

2022

Sensors

Self Cite

View full text Add to dashboard Cite

Deploying energy disaggregation models in the real-world is a challenging task. These models are usually deep neural networks and can be costly when running on a server or prohibitive when the target device has limited resources. Deep learning models are usually computationally expensive and they have large storage requirements. Reducing the computational cost and the size of a neural network, without trading off any performance is not a trivial task. This paper suggests a novel neural architecture that has less learning parameters, smaller size and fast inference time without trading off performance. The proposed architecture performs on par with two popular strong baseline models. The key characteristic is the Fourier transformation which has no learning parameters and it can be computed efficiently.

show abstract

Section: Introductionmentioning

confidence: 93%

Section: Related Workmentioning

confidence: 99%

Section: On House 1 From Uk-dalementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Neural Fourier Energy Disaggregation

Nalmpantis

Gkalinikis

Vrakas

2022

Sensors

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although the cost of measurement tends to reduce over time, minimising the cost and complexity of systems will be important for widespread adoption [60]. There is significant potential for the CPD methods described here to augment energy disaggregation systems that can identify the consumption of individual appliances from the incoming mains supply, where there has been recent developments based on artificial neural networks [61] that outperform existing Factorial Hidden Markov Models (FHMM). There is also advantage in context-related information based on low frequency data over the more typical high frequency methods due to the attendant reductions in sensor cost [62].…”

Section: Data Processing Frameworkmentioning

confidence: 99%

The application of quality control charts for identifying changes in time-series home energy data

Vivancos

Buswell

Cosar-Jorda

et al. 2020

Energy and Buildings

View full text Add to dashboard Cite

Energy consumption in the home is heavily influenced by the occupants and the routines they adopt. Although these routines tend to be regarded as somewhat static in nature, more recent evidence from the social sciences suggests that patterns of consumption are actually more fluid and constantly evolve to accommodate the contingencies of everyday living. This makes detecting changes in patterns of activity and their impact on energy consumption difficult, particularly when these patterns are often invisible to the householder to begin with. Being able to identify when a change occurs, therefore, could be a powerful tool to establish the context of change and so to determine more appropriate corrective action to curb waste and create opportunities for greater flexibility in consumption. The growing adoption of smart meters and home energy monitoring provide a platform for numerical approaches, yet there is little work reported in the literature and none that have attempted to evaluate effectiveness of such methods applied to detect changes in behaviour using field monitoring data from family homes. This paper reports on the application of a Change Point Detection method based on statistical quality control charts applied to identify changes in activities a family home using typical monitoring data. The approach was found to be very effective, identifying 78% of the changes that occurred over a two-year period and hence the outlook for such methods is promising. The findings suggest that such techniques could significantly improve the quality of information provided in energy feedback and so could play a significant role in the pursuit of more efficient energy use in the home by adding value to monitoring systems and services.

show abstract