No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Appliance consumption signature database and recognition test protocols
Abstract
We report on the creation of a database of appliance consumption signatures and two test protocols to be used for appliance recognition tasks. By means of plug-based low-end sensors measuring the electrical consumption at low frequency, typically every 10 seconds, we made two acquisition sessions of one hour on about 100 home appliances divided into 10 categories: mobile phones (via chargers), coffee machines, computer stations (including monitor), fridges and freezers, Hi-Fi systems (CD players), lamp (CFL), laptops (via chargers), microwave oven, printers, and televisions (LCD or LED). We measured their consumption in terms of real power (W), reactive power (var), RMS current (A) and phase of voltage relative to current (φ). We now give free access to this ACS-Fl database. The proposed test protocols will help the scientific community to objectively compare new algorithms.
... We evaluate our method DEMUX on seven popular real-world multi-class time series datasets: ACSF1 [31], Plane [32], Trace [33], Rock [34], SmallKitchen [32], ECG5000 [35], Meat [36]. Each is a popular publiclyavailable dataset [32] for multi-class time series classiőcation. ...
... Ablation study. To demonstrate the need for each component of DEMUX, we perform an ablation study, removing different DEMUX components and reporting the AUC-Difference on the ACSF1 dataset [31]. We focus on the replacement strategy, mask memory unit, and each loss component. ...
... Runtime Performance. To explore DEMUX's practicality, we perform a runtime comparison of DEMUX with the nine state-of-the-art comparisons on ACSF1 dataset [31] to generate explanations for a three-layered fully-connected network. Our results are shown in Figure 10. ...
Explainability helps users trust deep learning solutions for time series classification. However, existing explainability methods for multi-class time series classifiers focus on one class at a time, ignoring relationships between the classes. Instead, when a classifier is choosing between many classes, an effective explanation must show what sets the chosen class apart from the rest. We now formalize this notion, studying the open problem of class-specific explainability for deep time series classifiers, a challenging and impactful problem setting. We design a novel explainability method, DEMUX, which learns saliency maps for explaining deep multi-class time series classifiers by adaptively ensuring that its explanation spotlights the regions in an input time series that a model uses specifically to its predicted class. DEMUX adopts a gradient-based approach composed of three interdependent modules that combine to generate consistent, class-specific saliency maps that remain faithful to the classifier's behavior yet are easily understood by end users. We demonstrate that DEMUX outperforms nine state-of-the-art alternatives on seven popular datasets when explaining two types of deep time series classifiers. We analyze run time performance, show the impacts of hyperparameter selection, and introduce a detailed study of perturbation methods for time series. Further, through a case study, we demonstrate that DEMUX's explanations indeed highlight what separates the predicted class from the others in the eyes of the classifier.
... One strong force behind their rising importance is the increasing use of sensors for automatic and high resolution monitoring in domains like smart homes (Jerzak and Ziekow 2014), starlight observations (Protopapas et al. 2006), machine surveillance (Mutschler et al. 2013), or smart grids (Hobbs et al. 1999;Lew and Milligan 2016). Time series classification (TSC) is the problem of assigning one of a predefined class to a time series, like recognizing the electronic device producing a certain temporal pattern of energy consumption (Gao et al. 2014;Gisler et al. 2013) or classifying a signal of earth motions as either an earthquake or a passing lorry (Perol et al. 2018). ...
... In such settings, it is suboptimal to wait for a fixed number of measurements; instead, the algorithm should wait for the characteristic patterns, which may occur early (in which case an early classification is possible) or later (in which case the eTSC algorithm has to wait longer). As an example, Fig. 1 illustrates traces for the Fig. 1 Traces of microwaves taken from (Gisler et al. 2013). The operational state of the microwave starts between 5% and 50% of the whole trace length. ...
... The operational state of the microwave starts between 5% and 50% of the whole trace length. To have at least one event (typically a high burst of energy consumption) for each microwave, the threshold has to be set to that of the latest seen operational state (after seeing more than 46.2%) operational state of microwaves (Gisler et al. 2013). Observations started while the microwaves were already under power; the concrete operational state, characterized by high bursts of energy consumption, happened after 5-50% of the whole trace (amounting to 1 hour). ...
Early time series classification (eTSC) is the problem of classifying a time series after as few measurements as possible with the highest possible accuracy. The most critical issue of any eTSC method is to decide when enough data of a time series has been seen to take a decision: Waiting for more data points usually makes the classification problem easier but delays the time in which a classification is made; in contrast, earlier classification has to cope with less input data, often leading to inferior accuracy. The state-of-the-art eTSC methods compute a fixed optimal decision time assuming that every times series has the same defined start time (like turning on a machine). However, in many real-life applications measurements start at arbitrary times (like measuring heartbeats of a patient), implying that the best time for taking a decision varies widely between time series. We present TEASER, a novel algorithm that models eTSC as a two-tier classification problem: In the first tier, a classifier periodically assesses the incoming time series to compute class probabilities. However, these class probabilities are only used as output label if a second-tier classifier decides that the predicted label is reliable enough, which can happen after a different number of measurements. In an evaluation using 45 benchmark datasets, TEASER is two to three times earlier at predictions than its competitors while reaching the same or an even higher classification accuracy. We further show TEASER’s superior performance using real-life use cases, namely energy monitoring, and gait detection.
... One strong force behind their rising importance is the increasing use of sensors for automatic and high resolution monitoring in domains like smart homes [15], starlight observations [24], machine surveillance [20], or smart grids [14,32]. Time series classification (TSC) is the problem of assigning one of a predefined class to a time series, like recognizing the electronic device producing a certain temporal pattern of energy consumption [9,11] or classifying a signal of earth motions as either an earthquake or a bypassing lorry [23]. Conventional TSC works on time series of a given, fixed length and assumes access to the entire input at classification time. ...
... In contrast, early time series classification (eTSC), which we study in this work, tries to solve the TSC problem after seeing as few measurements as possible [34]. This need arises when the classification decision is time-critical, for instance to prevent damage (the earlier a warning system can predict an earthquake from seismic data [23], the more time there is for preparation), to speed-up diagnosis (the earlier an abnormal heart-beat is detected, the more time there is for prevention of fatal attacks [13]), or to protect markets and systems (the earlier 11.5% Total: 100% 8.5% 46.2% ...
... Predictions are only safe after this fraction of the data, as we have seen events from all traces Characteristic event Characteristic event Figure 1: Traces of microwaves taken from [11]. The operational state of the microwave starts between 5% and 50% of the whole trace length. ...
Early time series classification (eTSC) is the problem of classifying a time series after as few measurements as possible with the highest possible accuracy. The most critical issue of any eTSC method is to decide when enough data of a time series has been seen to take a decision: Waiting for more data points usually makes the classification problem easier but delays the time in which a classification is made; in contrast, earlier classification has to cope with less input data, often leading to inferior accuracy. The state-of-the-art eTSC methods compute a fixed optimal decision time assuming that every times series has the same defined start time (like turning on a machine). However, in many real-life applications measurements start at arbitrary times (like measuring heartbeats of a patient), implying that the best time for taking a decision varies heavily between time series. We present TEASER, a novel algorithm that models eTSC as a two two-tier classification problem: In the first tier, a classifier periodically assesses the incoming time series to compute class probabilities. However, these class probabilities are only used as output label if a second-tier classifier decides that the predicted label is reliable enough, which can happen after a different number of measurements. In an evaluation using 45 benchmark datasets, TEASER is two to three times earlier at predictions than its competitors while reaching the same or an even higher classification accuracy. We further show TEASER's superior performance using real-life use cases, namely energy monitoring, and gait detection.
... The resulting TSC problem is as follows: Given an energy consumption profile (which is a TS), determine the device type based on a set of exemplary profiles per type. For an energy company such information helps to improve the prediction of future energy consumption [12,13]. For approaching these kinds of problems, algorithms that are very fast and very accurate are required. ...
... Appliance load monitoring has become an important tool for energy savings [12,13]. We tested the performance of different TSC methods on data obtained from intrusive load monitoring (ILM), where energy consumption is separately recorded at every electric device. ...
... We tested the performance of different TSC methods on data obtained from intrusive load monitoring (ILM), where energy consumption is separately recorded at every electric device. We used two publicly available datasets ACS-F1 [13] and PLAID [12]. These datasets capture the power consumption of typical appliances. ...
Time series (TS) occur in many scientific and commercial applications, ranging from earth surveillance to industry automation to the smart grids. An important type of TS analysis is classification, which can, for instance, improve energy load forecasting in smart grids by detecting the types of electronic devices based on their energy consumption profiles recorded by automatic sensors. Such sensor-driven applications are very often characterized by (a) very long TS and (b) very large TS datasets needing classification. However, current methods to time series classification (TSC) cannot cope with such data volumes at acceptable accuracy; they are either scalable but offer only inferior classification quality, or they achieve state-of-the-art classification quality but cannot scale to large data volumes. In this paper, we present WEASEL (Word ExtrAction for time SEries cLassification), a novel TSC method which is both fast and accurate. Like other state-of-the-art TSC methods, WEASEL transforms time series into feature vectors, using a sliding-window approach, which are then analyzed through a machine learning classifier. The novelty of WEASEL lies in its specific method for deriving features, resulting in a much smaller yet much more discriminative feature set. On the popular UCR benchmark of 85 TS datasets, WEASEL is more accurate than the best current non-ensemble algorithms at orders-of-magnitude lower classification and training times, and it is almost as accurate as ensemble classifiers, whose computational complexity makes them inapplicable even for mid-size datasets. The outstanding robustness of WEASEL is also confirmed by experiments on two real smart grid datasets, where it out-of-the-box achieves almost the same accuracy as highly tuned, domain-specific methods.
... The resulting TSC problem is as follows: Given an energy consumption profile (which is a TS), determine the device type based on a set of exemplary profiles per type. For an energy company such information helps to improve the prediction of future energy consumption [14,13]. For approaching these kinds of problems, algorithms that are very fast and very accurate are required. ...
... We then search for the c split points that maximize the information gain (line 6). After choosing the first split point (line 10) any remaining partition YL or YR that is not pure is recursively split (lines [13][14]. The recursion ends once we have found c bins (line 12). ...
... Appliance load monitoring has become an important tool for energy savings [14,13]. We tested the performance of different TSC methods on data obtained from intrusive load monitoring (ILM), where energy consumption is separately recorded at every electric device. ...
Time series (TS) occur in many scientific and commercial applications, ranging from earth surveillance to industry automation to the smart grids. An important type of TS analysis is classification, which can, for instance, improve energy load forecasting in smart grids by detecting the types of electronic devices based on their energy consumption profiles recorded by automatic sensors. Such sensor-driven applications are very often characterized by (a) very long TS and (b) very large TS datasets needing classification. However, current methods to time series classification (TSC) cannot cope with such data volumes at acceptable accuracy; they are either scalable but offer only inferior classification quality, or they achieve state-of-the-art classification quality but cannot scale to large data volumes. In this paper, we present WEASEL (Word ExtrAction for time SEries cLassification), a novel TSC method which is both scalable and accurate. Like other state-of-the-art TSC methods, WEASEL transforms time series into feature vectors, using a sliding-window approach, which are then analyzed through a machine learning classifier. The novelty of WEASEL lies in its specific method for deriving features, resulting in a much smaller yet much more discriminative feature set. On the popular UCR benchmark of 85 TS datasets, WEASEL is more accurate than the best current non-ensemble algorithms at orders-of-magnitude lower classification and training times, and it is almost as accurate as ensemble classifiers, whose computational complexity makes them inapplicable even for mid-size datasets. The outstanding robustness of WEASEL is also confirmed by experiments on two real smart grid datasets, where it out-of-the-box achieves almost the same accuracy as highly tuned, domain-specific methods.
... Ablation study To demonstrate the need for each component of DEMUX, we perform an ablation study, removing different DEMUX components and reporting the AUC-Difference on the ACSF1 dataset [31]. We focus on the replacement strategy, mask memory unit, and each loss component. ...
... Additionally, for a fair comparison with DEMUX results, the hyperparameters of the compared methods (DYNA, PERT, TMULE, TSHAP, SHAP, LEFT, LIME, and MP) are tuned for ACSF1 dataset, according to the explanations provided in their respective papers [6-9, 11-13, 19, 40]. Runtime performance To explore DEMUX's practicality, we perform a runtime comparison of DEMUX with the nine state-of-the-art comparisons on ACSF1 dataset [31] to generate explanations for a three-layered fully-connected network. Our results are shown in Fig. 10. ...
Explainability helps users trust deep learning solutions for time series classification. However, existing explainability methods for multi-class time series classifiers focus on one class at a time, ignoring relationships between the classes. Instead, when a classifier is choosing between many classes, an effective explanation must show what sets the chosen class apart from the rest. We now formalize this notion, studying the open problem of class-specific explainability for deep time series classifiers, a challenging and impactful problem setting. We design a novel explainability method, DEMUX, which learns saliency maps for explaining deep multi-class time series classifiers by adaptively ensuring that its explanation spotlights the regions in an input time series that a model uses specifically to its predicted class. DEMUX adopts a gradient-based approach composed of three interdependent modules that combine to generate consistent, class-specific saliency maps that remain faithful to the classifier’s behavior yet are easily understood by end users. We demonstrate that DEMUX outperforms nine state-of-the-art alternatives on seven popular datasets when explaining two types of deep time series classifiers. We analyze runtime performance, show the impacts of hyperparameter selection, and introduce a detailed study of perturbation methods for time series. Further, through a case study, we demonstrate that DEMUX’s explanations indeed highlight what separates the predicted class from the others in the eyes of the classifier.
... We evaluate our method DEMUX on five popular real-world multi-class time series datasets: ACSF1 [29], PLANE [30], TRACE [31], ROCK [32], ECG5000 [33], MEAT [34]. Each is a popular publicly-available dataset [30] for multi-class time series classification. ...
... To demonstrate the need for each component of DEMUX, we perform an ablation study, removing different DEMUX components and reporting the AUC-Difference on the ACSF1 dataset [29]. We focus on the replacement strategy, mask memory unit, and each loss component. ...
Explainability helps users trust deep learning solutions for time series classification. However, existing explainability methods for multi-class time series classifiers focus on one class at a time, ignoring relationships between the classes. Instead, when a classifier is choosing between many classes, an effective explanation must show what sets the chosen class apart from the rest. We now formalize this notion, studying the open problem of class-specific explainability for deep time series classifiers, a challenging and impactful problem setting. We design a novel explainability method, DEMUX, which learns saliency maps for explaining deep multi-class time series classifiers by adaptively ensuring that its explanation spotlights the regions in an input time series that a model uses specifically to its predicted class. DEMUX adopts a gradient-based approach composed of three interdependent modules that combine to generate consistent, class-specific saliency maps that remain faithful to the classifier's behavior yet are easily understood by end users. Our experimental study demonstrates that DEMUX outperforms nine state-of-the-art alternatives on five popular datasets when explaining two types of deep time series classifiers. Further, through a case study, we demonstrate that DEMUX's explanations indeed highlight what separates the predicted class from the others in the eyes of the classifier. Our code is publicly available at https://github.com/rameshdoddaiah/DEMUX.
... Steady-state features are mainly active and reactive power and a set of statistical low-frequency features computed from the active or reactive power (e.g. mean, median, variance or energy) [81], while also dimensionality reduction techniques like principal component analysis have been used to extract lower dimensional feature vectors and improve run time [82]. What regards transient features, they include current harmonics, Total Harmonic Distortion (THD) and transient energy [11], [40]. ...
... Table II provides a list of 29 databases with energy/power recordings from eleven countries, different sampling frequencies, number, and types of devices and monitoring duration lengths. Out of these 29 databases 18 databases (REDD [121], BLUED [122], ECO [123], UK-DALE [124], Dataport [125], Smart [126], RAE [127], iAWE [128], IHEPCDC [129], REFIT [130], AMPds [44], [131], COMBED [132], DRED [133], SustDataED [134], EEUD [135], SysD [136], LIFTED [137] and BLOND [115]) can be used for training NILM systems as they include both aggregated power consumption as well as consumption on device level, while five of them (RBSAM [138], HES [139], Tracebace [140], GREEND [141], ACS-F1/2 [81]) only include power consumption on appliance level hence are not suitable for testing NILM approaches. Furthermore, there is a set of six (PLAID [142], WHITED [143], LILAC [144], CREAM [145], HFED [146], COOLL [147]) additional databases which consist of transient appliance signatures and can only be used for extracting features, create transient appliance models or design edge detectors. ...
The rapid development of technology in the elec-
trical energy sector within the last 20 years has led to grow-
ing electric power needs through the increased number of
electrical appliances and automation of tasks. In parallel the
global climate protection goals, energy conservation and efficient
energy management arise interest for reduction of the overall
energy consumption. These requirements have led to the recent
adoption of smart-meters and smart-grids, as well as to the
rise of Load Monitoring (LM) using energy disaggregation, also
referred to as Non-Intrusive Load Monitoring (NILM), which
enables appliance-specific energy monitoring by only observing
the aggregated energy consumption of a household. The real-time
information on appliance level can be used to get deeper insights
in the origin of energy consumption and to make optimization,
strategic load scheduling and demand management feasible. The
three main contributions are as follows: First, a generalized
up-to-date review of NILM approaches including a high-level
taxonomy of NILM methodologies is provided. Second, previously
published results are grouped based on the experimental setup
which allows direct comparison. Third, the article is accompanied
by a software implementation of the described NILM approaches.
... Steady-State features are mainly active and reactive power and a set of statistical lowfrequency features computed from the active or reactive power (e.g. mean, median, variance or energy) [74]. Contrary, transient features are limited by the sampling frequency of the respective database and features, e.g. ...
... In detail the columns in Table 2.3 list the year and country in which the database was recorded, the number of houses and devices '#-houses/devices', the monitoring duration, the measurement approach, the measured features and the sampling resolution separately for device level and aggregated data. [80], BLUED [84], ECO [85], UK-DALE [86], Dataport [87], Smart [88], RAE [89], iAWE [90], IHEPCDC [91], REFIT [92], AMPd [43], COMBED [93], DRED [94], SustDataED [95], EEUD [96] and BLOND [97]) can be used for training state-based NILM systems as they include both aggregated power consumption as well as consumption on device level, while five of them (RBSAM [98], HES [99], Tracebace [100], GREEND [101], ACS-F1/2 [74]) only include power consumption on appliance level hence are not suitable for testing NILM approaches. Furthermore there is a set of six (PLAID [102], WHITED [103], LILAC [104], CREAM [105], HFED [106], COOLL ...
The rapid development of technology in the electrical sector within the last 20 years has led to growing electric power needs through the increased number of electrical appliances and automation of tasks. In contrary, reduction of the overall energy consumption as well as efficient energy management are needed, in order to reduce global warming and meet the global climate protection goals. These requirements have led to the recent adoption of smart-meters and smart-grids, as well as to the rise of Non-Intrusive Load Monitoring. Non-Intrusive Load Monitoring aims to extract the energy consumption of individual electrical appliances through disaggregation of the total power consumption as measured by a single smart meter at the inlet of a household. Therefore, Non-Intrusive Load Monitoring is a highly under-determined problem which aims to estimate multiple variables from a single observation, thus is impossible to be solved analytical. In order to find accurate estimates of the unknown variables three fundamentally different approaches, namely deep-learning, pattern matching and single-channel source separation, have been investigated in the literature in order to solve the Non-Intrusive Load Monitoring problem. While Non-Intrusive Load Monitoring has multiple areas of application, including energy reduction through consumer awareness, load scheduling for energy cost optimization or reduction of peak demands, the focus of this thesis is especially on the performance of the disaggregation algorithm, the key part of the Non-Intrusive Load Monitoring architecture. In detail, optimizations are proposed for all three architectures, while the focus lies on deep-learning based approaches. Furthermore, the transferability capability of the deep-learning based approach is investigated and a NILM specific transfer architecture is proposed. The main contribution of the thesis is threefold. First, with Non-Intrusive Load Monitoring being a time-series problem incorporation of temporal information is crucial for accurate modelling of the appliance signatures and the change of signatures over time. Therefore, previously published architectures based on deep-learning have focused on utilizing regression models which intrinsically incorporating temporal information. In this work, the idea of incorporating temporal information is extended especially through modelling temporal patterns of appliances not only in the regression stage, but also in the input feature vector, i.e. by using fractional calculus, feature concatenation or high-frequency double Fourier integral signatures. Additionally, multi variance matching is utilized for Non-Intrusive Load Monitoring in order to have additional degrees of freedom for a pattern matching based solution. Second, with Non-Intrusive Load Monitoring systems expected to operate in realtime as well as being low-cost applications, computational complexity as well as storage limitations must be considered. Therefore, in this thesis an approximation for frequency domain features is presented in order to account for a reduction in computational complexity. Furthermore, investigations of reduced sampling frequencies and their impact on disaggregation performance has been evaluated. Additionally, different elastic matching techniques have been compared in order to account for reduction of training times and utilization of models without trainable parameters. Third, in order to fully utilize Non-Intrusive Load Monitoring techniques accurate transfer models, i.e. models which are trained on one data domain and tested on a different data domain, are needed. In this context it is crucial to transfer time-variant and manufacturer dependent appliance signatures to manufacturer invariant signatures, in order to assure accurate transfer modelling. Therefore, a transfer learning architecture specifically adapted to the needs of Non-Intrusive Load Monitoring is presented. Overall, this thesis contributes to the topic of Non-Intrusive Load Monitoring improving the performance of the disaggregation stage while comparing three fundamentally different approaches for the disaggregation problem.
... Imagine that some process in the system we are monitoring occasionally "injects" a pattern into the time series. As a concrete example, washing machines typically have a prototypic signature (as exhibited in the TRACE dataset [40]), but the signatures express themselves more slowly on a cold day, when it takes longer to heat the cooler water supplied from the city [9]. We would like all equal length instances of the signature to have approximately the same distance. ...
... It seems that if the nearest neighbor of T i, ℓ min is T j, ℓ min , then probably the nearest neighbor of T i, ℓ min +1 is T j, ℓ min +1 . For example, as shown in Figure 3(bottom), if we sort the distance profiles of T 33,8 and T 33,9 in ascending order, we can find that the nearest neighbor of T 33,8 is T 97,8 , and the nearest neighbor of T 33,9 is T 97,9 . ...
In the last fifteen years, data series motif discovery has emerged as one of the most useful primitives for data series mining, with applications to many domains, including robotics, entomology, seismology, medicine, and climatology. Nevertheless, the state-of-the-art motif discovery tools still require the user to provide the motif length. Yet, in at least some cases, the choice of motif length is critical and unforgiving. Unfortunately, the obvious brute-force solution, which tests all lengths within a given range, is computationally untenable. In this work, we introduce VALMOD, an exact and scalable motif discovery algorithm that efficiently finds all motifs in a given range of lengths. We evaluate our approach with five diverse real datasets, and demonstrate that it is up to 20 times faster than the state-of-the-art. Our results also show that removing the unrealistic assumption that the user knows the correct length, can often produce more intuitive and actionable results, which could have been missed otherwise.
... The second method injects real electrical appliance signatures into the demand. The libraries of electrical signature provide various electricity signatures, such as a television, microwave oven, and printer [28,29]. The last method replays the occupied power usage of a few days ago. ...
... The database includes about 100 home appliances divided into 10 categories, such as coffee machines, computer stations (including monitors), and microwave ovens. Specific details of the database can be found in [29,33]. ...
Occupancy information is one of the most important privacy issues of a home. Unfortunately, an attacker is able to detect occupancy from smart meter data. The current battery-based load hiding (BLH) methods cannot solve this problem. To thwart occupancy detection attacks, we propose a framework of battery-based schemes to prevent occupancy detection (BPOD). BPOD monitors the power consumption of a home and detects the occupancy in real time. According to the detection result, BPOD modifies those statistical metrics of power consumption, which highly correlate with the occupancy by charging or discharging a battery, creating a delusion that the home is always occupied. We evaluate BPOD in a simulation using several real-world smart meter datasets. Our experiment results show that BPOD effectively prevents the threshold-based and classifier-based occupancy detection attacks. Furthermore, BPOD is also able to prevent nonintrusive appliance load monitoring attacks (NILM) as a side-effect of thwarting detection attacks.
... Source: G. Hart Patent [4] The ACS-F2 dataset was selected for this work because of its low sampling rate which is similar to the commercial energy meters and could be easily deployed after the parameters are finalized [8]. This dataset, an improved version of the previously developed ACS-F1 dataset [9] comprises 15 classes, each representing a different appliance, and 15 subclasses for each class representing different manufacturing brands, totaling approximately 225 load signatures of appliances. This dataset consists of two distinct instances: A1 and A2 used for training and testing respectively. ...
The Energy Sector across the globe is experiencing rapid growth, driven by Internet of Things (IoT) integration technologies and advanced algorithms. This evolution is particularly evident in the ongoing competition among tech companies in the development of smart metering solutions. Despite these advancements, a critical challenge persists—the lack of definitive technical protocols for monitoring the total usage or power signatures of individual appliances, referred to as non-intrusive load monitoring (NILM) in aggregate. While intrusive load monitoring (ILM) provides very accurate and thorough insights, non-intrusive methods are essential to address losses specially in residential areas. In this research a groundbreaking approach is proposed towards handling NILM problems by analyzing and aggregating the load patterns of four key appliances of daily use, namely the Coffee Machine, Fridge, Kettle, and Laptop from the ACS-F2 dataset. The generated aggregated dataset, is systematically combined using electrical formulations to yield the desired data which reflects the simultaneous operation of multiple appliances, this has been explored for the first time in the known literature. The proposed dataset contains around 6750 aggregated appliance load patterns for both training and testing. Furthermore, multiple Time Series Classifiers (TSC) were gauged using a suite of evaluation metrics, on the proposed dataset and an accuracy of 92.1% was achieved by the CATCH22 classifier.
... Low-frequency datasets are more available to the public due to the lower cost and easier installation. The sampling rate and features vary from one dataset to another where the sampling rate in the datasets is between 0. [42] datasets recorded the consumption of the appliance at a sampling rate of 10 seconds as well commercial building energy dataset (COMBED) [43] dataset collected data at 20 seconds. There are datasets having resolution of 1 minutes such as almanac of minutely power dataset (AMPDs2) [44], Chulalongkorn university building energy management system (CU-BEMS) [45], Indian buildings energy consumption dataset (I-BLEND) [46], electrical-end-use dataset (EEUD) [47], and Dutch residential energy dataset (DRED) [48]. ...
The implementation of non-intrusive load monitoring has gained significant attention as a promising solution for disaggregating and identifying individual appliances' energy consumption within households and commercial buildings. The issue at the core of non-intrusive load monitoring is the accurate separation of the aggregated energy consumption signal into the energy contributions of distinct appliances without requiring additional sensors or invasive measurements. The erroneous implementation of non-intrusive load monitoring model leads to poor results and a waste of time where the criteria for the implementation are necessary. This paper presents an overview of non-intrusive load monitoring's implementation, addressing the stages of non-intrusive load monitoring and advancements in the field. The overview explores the various techniques, methodologies, and algorithms employed in non-intrusive load monitoring systems to achieve load identification and energy disaggregation. By addressing the problem statement and reviewing the state of the art, this paper sets the stage for understanding the significance of non-intrusive load monitoring in commercial buildings and modern energy-efficient households. Moreover, the limitations, applications, and recommended criteria for the implementation of non-intrusive load monitoring are highlighted as a contribution to the field of load disaggregation. Future research should address scalability, privacy concerns, and the integration of non-intrusive load monitoring with smart grids to propel energy disaggregation towards more efficient, accessible, and sustainable energy management systems. We envision that this endeavour can serve as a guiding beacon, offering a roadmap for researchers and professionals.
... The ACS-F1 database is a dataset of electricity consumption for 10 categories of household appliances: mobile phones (via chargers), coffee machines, computer stations (including monitors), fridges and freezers, Hi-Fi systems (CD players), lamp (CFL), laptops (via chargers), microwave ovens, printers, and televisions (LCD or LED), measured at low frequencies (usually every 10 s) via plug-based low-end sensors. Energy load forecasting in smart grids can be improved by determining the type of electronic equipment based on energy consumption profiles recorded by automatic sensors [45,46]. Very long time series and very large time series datasets requiring classification are characteristic of such sensor-driven applications. ...
In this paper, we propose a multidimensional scaling (MDS) method based on the Wasserstein–Fourier (WF) distance to analyze and classify complex time series from a frequency domain perspective in complex systems. Three properties with rigorous derivation are stated to reveal the basics structure of MDS method based on the WF distance and validate it as an excellent metric for time series classification. Our proposed method solves the problem of non-equal-length sequence distances faced by traditional metrics and the problem that the Kullback–Leibler divergence and Jensen–Shannon divergence cannot measure completely non-overlapping distributions. Its practicability for multiple data sets and robustness to noise are demonstrated in the paper. Compared with the MDS methods based on Euclidean distance, Chebyshev distance, correlation distance, complexity-invariant distance and dynamic time warping, our proposed method is more effective and has minimal losses and errors.
... Database of appliance consumption signatures (ACS-F1 and ACS-F2) [26,27]: It provides IL electricity consumption data taken from hundreds of Swiss houses. There is a 2013 release (called ACS-F1) and a 2014 release (called ACS-F2). ...
This paper consists of two parts: an overview of existing open datasets of electricity consumption and a description of the Moroccan Buildings’ Electricity Consumption Dataset, a first of its kind, coined as MORED. The new dataset comprises electricity consumption data of various Moroccan premises. Unlike existing datasets, MORED provides three main data components: whole premises (WP) electricity consumption, individual load (IL) ground-truth consumption, and fully labeled IL signatures, from affluent and disadvantaged neighborhoods. The WP consumption data were acquired at low rates (1/5 or 1/10 samples/s) from 12 households; the IL ground-truth data were acquired at similar rates from five households for extended durations; and IL signature data were acquired at high and low rates (50 k and 4 samples/s) from 37 different residential and industrial loads. In addition, the dataset encompasses non-intrusive load monitoring (NILM) metadata.
... The ACS-Fx 11 , 12 [31] contains electricity consumption signatures for appliance recognition tasks with the help of two test protocols. Plug-based sensors measured the electrical consumption at low frequency, typically every 10 seconds. ...
Nowadays Non-Intrusive Load Monitoring (NILM) is considered a hot topic among researchers. The energy disaggregation datasets are used as the benchmark to validate the performance of energy disaggregation algorithms. It is indeed rather difficult to record the load monitoring of devices and appliances; therefore various benchmarking datasets have been proposed during the past few years. This paper presentsa comprehensive review of 42 NILM datasets aided by comparison tables, generated to elaborate on the diverse features of existing datasets. Moreover, the strengths and limitations of present NILM datasets are highlighted with an outlook on present challenges and future research directions as a contribution to the field of energy disaggregation and load identification. The review will help the researchers to evaluate the performance of new NILM algorithms. We believe that this work could be served as a guideline and can potentially open new research perspectives to the scientific community working on developing new NILM datasets.
... ACS-F1 [66] and BERDS [67] datasets that monitor load patterns at a comparable sampling rate are proposed. ACS-F1 records the amount of energy used in a set of households at an appliancelevel. ...
In the last decade, extended efforts have been poured into energy efficiency. Several energy consumption datasets were henceforth published, with each dataset varying in properties, uses and limitations. For instance, building energy consumption patterns are sourced from several sources, including ambient conditions, user occupancy, weather conditions and consumer preferences. Thus, a proper understanding of the available datasets will result in a strong basis for improving energy efficiency. Starting from the necessity of a comprehensive review of existing databases, this work is proposed to survey, study and visualize the numerical and methodological nature of building energy consumption datasets. A total of thirty-one databases are examined and compared in terms of several features, such as the geographical location, period of collection, number of monitored households, sampling rate of collected data, number of sub-metered appliances, extracted features and release date. Furthermore, data collection platforms and related modules for data transmission, data storage and privacy concerns used in different datasets are also analyzed and compared. Based on the analytical study, a novel dataset has been presented, namely Qatar university dataset, which is an annotated power consumption anomaly detection dataset. The latter will be very useful for testing and training anomaly detection algorithms, and hence reducing wasted energy. Moving forward, a set of recommendations is derived to improve datasets collection, such as the adoption of multi-modal data collection, smart Internet of things data collection, low-cost hardware platforms and privacy and security mechanisms. In addition, future directions to improve datasets exploitation and utilization are identified, including the use of novel machine learning solutions, innovative visualization tools and explainable mobile recommender systems. Accordingly, a novel visualization strategy based on using power consumption micro-moments has been presented along with an example of deploying machine learning algorithms to classify the micro-moment classes and identify anomalous power usage.
... ACS-F1 [66] and BERDS [67] datasets that monitor load patterns at a comparable sampling rate are proposed. ACS-F1 records the amount of energy used in a set of households at an appliancelevel. ...
In the last decade, extended efforts have been poured into energy efficiency. Several energy consumption datasets were henceforth published, with each dataset varying in properties, uses and limitations. For instance, building energy consumption patterns are sourced from several sources, including ambient conditions, user occupancy, weather conditions and consumer preferences. Thus, a proper understanding of the available datasets will result in a strong basis for improving energy efficiency. Starting from the necessity of a comprehensive review of existing databases, this work is proposed to survey, study and visualize the numerical and methodological nature of building energy consumption datasets. A total of thirty-one databases are examined and compared in terms of several features, such as the geographical location, period of collection, number of monitored households, sampling rate of collected data, number of sub-metered appliances, extracted features and release date. Furthermore, data collection platforms and related modules for data transmission, data storage and privacy concerns used in different datasets are also analyzed and compared. Based on the analytical study, a novel dataset has been presented, namely Qatar university dataset, which is an annotated power consumption anomaly detection dataset. The latter will be very useful for testing and training anomaly detection algorithms, and hence reducing wasted energy. Moving forward, a set of recommendations is derived to improve datasets collection, such as the adoption of multi-modal data collection, smart Internet of things data collection, low-cost hardware platforms and privacy and security mechanisms. In addition, future directions to improve datasets exploitation and utilization are identified, including the use of novel machine learning solutions, innovative visualization tools and explainable recommender systems.
... Imagine that some process in the system we are monitoring occasionally "injects" a pattern into the time series. As a concrete example, washing machines typically have a prototypic signature (as exhibited in the TRACE dataset (Roverso, 2000)), but the signatures express themselves more slowly on a cold day, when it takes longer to heat the cooler water supplied from the city (Gisler et al., 2013). We would like all equal length instances of the signature to have approximately the same distance. ...
In the last fifteen years, data series motif and discord discovery have emerged as two useful and well-used primitives for data series mining, with applications to many domains, including robotics, entomology, seismology, medicine, and climatology. Nevertheless, the state-of-the-art motif and discord discovery tools still require the user to provide the relative length. Yet, in several cases, the choice of length is critical and unforgiving. Unfortunately, the obvious brute-force solution, which tests all lengths within a given range, is computationally untenable. In this work, we introduce a new framework, which provides an exact and scalable motif and discord discovery algorithm that efficiently finds all motifs and discords in a given range of lengths. We evaluate our approach with five diverse real datasets, and demonstrate that it is up to 20 times faster than the state-of-the-art. Our results also show that removing the unrealistic assumption that the user knows the correct length, can often produce more intuitive and actionable results, which could have otherwise been missed. (Paper published in Data Mining and Knowledge Discovery Journal - 2020)
... There have been also been recent developments in the application of DL for energy prediction, with promising results showing good prediction and robustness. One factor contributing to these developments is the availability of relevant data [29][30][31]. Google recently implemented a deep neural network (DNN) with five hidden layers and 50 nodes per hidden layer to increase power user efficiency (PUE) in their data centers [32]. ...
The U.S. power grid is transforming to become “smarter,” cleaner, and more efficient. This is leading to the addition of significant distributed variable renewable generation. Due to the variable nature of renewable generation, the short- and long-term supply-demand imbalances are less predictable, and conventional approaches to mitigating the imbalance will not be efficient or cost-effective. To address this challenge, transactive control technologies have been proposed. The transactive control approach requires individual end-use loads to express flexibility as a function of price. To model flexibility while maintaining robustness to any non-linear behavior exhibited by end-use loads, machine learning approaches for load forecasting are being explored. However, certain aspects, such as how much training data is required and how deep models for load forecasting should be structured and trained are not well understood. This work explores how to apply sequence-to-sequence recurrent neural networks to short-term electrical load forecasting with a case study of four commercial office buildings. We find that it is best to start the training in the middle of a heating or cooling season with at least six months of data. We further show that models perform best when predictions are conditioned on three to 12 hours of prior data, with a decrease in performance for shorter contexts. We identify recommended ranges for common hyperparameters that could be used by practitioners applying similar models to their own tasks. Finally, we find that transferability of models across buildings is highly dependent on the building pairs, but in the best case, models are highly transferable.
... Ridi et al. [18] also conducted several plug load identification studies using the Appliance Consumption Signature-Fribourg 2 (ACS-F2) database. The ACS-F2 database contains the power signatures of 225 plug loads from 15 different categories and is an extension of the ACS-F1 database, which was released one year earlier in 2013 [19]. The power consumption information of each plug load is recorded at a sampling frequency of 1/10 Hz (equivalent to 1 sample every 10 seconds) using smart power plugs for one hour per session, over two separate sessions. ...
Plug loads account for up to one-third of the overall energy use in commercial buildings. There is thus a growing research interest in utilising load monitoring systems to track plug load usage by installing smart plugs to capture high-resolution consumption data. The availability of such data has also enabled the development of automatic plug load identification models that enhance the capabilities of existing load monitoring systems. Through our literature review, we highlighted several limitations that impede real-world implementation, such as the limited number of publicly available datasets for commercial buildings, models trained on data with high sampling frequencies while using an extended time window, and data leakage issues during model training. In this study, we proposed a near-real-time plug load identification approach that uses low-frequency power data (1/60 Hz) to identify plug loads in office spaces. The dataset used in this study is processed by first identifying the active periods of the plug loads before applying a novel dynamic time window strategy during feature extraction. These extracted features are subsequently passed through several classification algorithms and evaluated using different accuracy metrics. The proposed approach is also assessed through multiple experiments, including (1) identifying the best online and offline models, (2) comparing between different time window strategies, and (3) evaluating model performances under different sampling frequencies. As a result, the best online model achieved accuracies up to 93% using the Bagging algorithm with a minimum dynamic time window of 5 minutes. Finally, we highlighted two application areas of automatic plug load identification in energy dashboards and personalised control systems as part of future works.
... Imagine that some process in the system we are monitoring occasionally "injects" a pattern into the time series. As a concrete example, washing machines typically have a prototypic signature (as exhibited in the TRACE dataset (Roverso 2000)), but the signatures express themselves more slowly on a cold day, when it takes longer to heat the cooler water supplied from the city (Gisler et al. 2013). We would like all equal length instances of the signature to have approximately the same distance. ...
In the last 15 years, data series motif and discord discovery have emerged as two useful and well-used primitives for data series mining, with applications to many domains, including robotics, entomology, seismology, medicine, and climatology. Nevertheless, the state-of-the-art motif and discord discovery tools still require the user to provide the relative length. Yet, in several cases, the choice of length is critical and unforgiving. Unfortunately, the obvious brute-force solution, which tests all lengths within a given range, is computationally untenable. In this work, we introduce a new framework, which provides an exact and scalable motif and discord discovery algorithm that efficiently finds all motifs and discords in a given range of lengths. We evaluate our approach with five diverse real datasets, and demonstrate that it is up to 20 times faster than the state-of-the-art. Our results also show that removing the unrealistic assumption that the user knows the correct length, can often produce more intuitive and actionable results, which could have otherwise been missed.
... Common for all NILM approaches is that they use measurements of the aggregated energy consumption of a household with a sampling frequency f s in the order of a sample per second up to few tens of kHz [16]. NILM methods may use macroscopic signal parameters (e.g., active/reactive power [24,25]) or microscopic ones (e.g., transient energy and harmonics [26-28]) ...
A data-driven methodology to improve the energy disaggregation accuracy during Non-Intrusive Load Monitoring is proposed. In detail, the method is using a two-stage classification scheme, with the first stage consisting of classification models processing the aggregated signal in parallel and each of them producing a binary device detection score, and the second stage consisting of fusion regression models for estimating the power consumption for each of the electrical appliances. The accuracy of the proposed approach was tested on three datasets (ECO, REDD and iAWE), which are available online, using four different classifiers. The presented approach improves the estimation accuracy by up to 4.1% with respect to a basic energy disaggregation architecture, while the improvement on device level was up to 10.1%. Analysis on device level showed significant improvement of power consumption estimation accuracy especially for continuous and non-linear appliances across all evaluated datasets.
... • ACS-F1 ACS-F1 (Appliance Consumption Signatures -Fribourg 1) dataset [78] includes the electrical consumption measurement data (real power, reactive power, RMS current and phase of voltage relative to current (φ)) obtained through two acquisition sessions of one hour on about 100 home appliances. The measurement shave been conducted using plug-based sensors at low frequency (typically every 10 seconds) and the monitored devices include 10 categories of mobile phones, coffee machines, computer stations, fridges and freezers, CD players, lamps, laptops, microwave oven, printers, and televisions. ...
HYBRID wireless/powerline communication has been proposed to improve reliability
and performance in smart grids. In this work we focus on a hybrid wireless/
narrowband PLC (WiNPLC) system for communication links in Smart
Grids (SG) such as the link between smart meters (SM) and in-home devices (IHD). An
effective approach to enhance the data rate in narrowband power line communication
(NB-PLC) system is multicarrier modulation based on orthogonal frequency-division
multiplexing (OFDM) and multiple-input multiple-output (MIMO) transmission over
multiple power line phases.
Low voltage, in-home NB-PLC networks allow direct communication between SM
and IHD. In order to minimize security issues, in many deployment scenarios transmission
takes place only towards the IHD to display consumption data, with no backwards
channel. As a result, channel estimation is difficult and a key challenge for achieving
reliable communication over NB-PLC is to use robust transmission and detection techniques
to mitigate effect of the impulsive noise within the PLC channel and recover
transmitted data. To this aim, it is fundamental to give an appropriate characterization
of such a noise. In fact, substantial components of the noise in NB-PLC systems
exhibit a cyclostationary behavior with a period of half the alternating current (AC)
cycle. Moreover, when MIMO transmission is adopted, an important issue that must be
considered is the cross-correlation between different phases.
Frequency-shift (FRESH) filter is a recently proposed approach that is able to reproduce
the effects of cyclostationary NB-PLC noise obtained from measurements.
In this work, we propose to classify the noise generated by FRESH filter into three
classes, based on the estimation of the respective probability density functions and
on the evaluation of their second order statistics. First, we show that while one class
exhibits a normal distribution, the other two exhibit an impulsive behavior for which
we propose a generalized Student’s t-distribution. Simulation results show that the bit
error rate (BER) of MIMO-OFDM NB-PLC significantly changes between different
classes of noise. Hence, we develop an algorithm for switching data delivery between
MIMO-OFDM NB-PLC and MIMO-OFDM wireless transmission in unlicensed frequency
band that takes into account knowledge of the periodicity of the three classes of noises. The result is a hybrid MIMO-OFDM wireless/NB-PLC system, which we
refer to as, hybrid MIMO-OFDM WiNPLC. Our simulation results demonstrate BER
improvement of the proposed hybrid system over individual MIMO-OFDM NB-PLC
or MIMO-OFDM wireless systems. Further improvement in performance of hybrid
system could be obtained by evaluating capacity of the MIMO NB-PLC system in presence
of different classes of the noise. This thesis obtained capacity by spatio-temporal
whitening of the cyclostationary correlated noise samples generated through FRESH
filtering. This capacity is useful for adapting the modulation order and obtaining optimum
performance based on the class of noise. Due to the cyclostationarity of the
noise, similar behavior is repeated in next periods and can take advantages of this preprocessing.
To support the future works of other researchers in the field of NB-PLC, we propose
a simple and more adaptive method to generate noise samples with characteristics similar
to those obtained using the FRESH filter. To this aim, filterbank of FRESH filter is
replaced by applying spectral and temporal shaping to a white Gaussian (WGN) noise
random process to obtain correlated impulsive noise samples. In addition, by changing
the slope of temporal shaping, distribution of each class could change from Gaussian
to impulsive and vice versa. The proposed noise generation approach is compared with
FRESH filter generator in terms of normalized mean square error (NMSE) in the cyclic
auto-correlation, and bit error rate (BER), between the measured and generated noise
samples.
The noise introduced by electrical appliances to the communication data could be
used as device signatures which is an useful information for energy monitoring. In this
regard, data received to all SMs must be collected and analyzed for improving energy
consumption management. Due to the notable rise in the number of installed SMs,
Non-Intrusive Load Monitoring (NILM) has received growing attention in the recent
years. NILM aims at replacing several SMs by a single SM and estimating the power
demand of individual appliances from a household’s aggregate electricity consumption.
In the present work, after reviewing different categories of household appliances, the
state-of-the-art load signatures, including both macroscopic and microscopic features,
are introduced. Next, commonly used supervised and unsupervised disaggregation algorithms,
which are employed to classify the appliances based on the extracted features,
are discussed. Publically accessible datasets and open-source tools, which have been released
in the recent years to assist the NILM research and to facilitate the comparison of
disaggregation algorithms, are then reviewed. Finally, main applications of energy disaggregation,
including providing itemized energy bills, enabling more accurate demand
prediction, identifying malfunctioning appliances, and assisting occupancy monitoring,
are presented.
... Macroscopic features are mainly active and reactive power, while statistical values from the active or reactive power (e.g. mean, median, variance or energy) can be estimated as well [34]. Microscopic features can be current harmonics or transient energy [21,35] and require high-sampling frequency to be calculated (1 kHz and above). ...
Abstract An extension of the baseline non-intrusive load monitoring approach for energy disaggregation using temporal contextual information is presented in this paper. In detail, the proposed approach uses a two-stage disaggregation methodology with appliance-specific temporal contextual information in order to capture time-varying power consumption patterns in low-frequency datasets. The proposed methodology was evaluated using datasets of different sampling frequency, number and type of appliances. When employing appliance-specific temporal contextual information, an improvement of 1.5% up to 7.3% was observed. With the two-stage disaggregation architecture and using appliance-specific temporal contextual information, the overall energy disaggregation accuracy was further improved across all evaluated datasets with the maximum observed improvement, in terms of absolute increase of accuracy, being equal to 6.8%, thus resulting in a maximum total energy disaggregation accuracy improvement equal to 10.0%.
... Depending on the sampling frequency value N ILM methods utilize either macroscopic features (e.g. active/reactive power [11]) or microscopic features (e.g. transient energy, harmonics [12]- [14]) to disaggregate energy consumption into device level [15]. ...
... The features extracted from the aggregated signal in approaches with and without SS strongly depend on the sampling frequency, with either macroscopic (for low sampling frequency) or microscopic (for high sampling frequency) features being extracted. Macroscopic features are mainly active and reactive power, while statistical values from the active or reactive power (e.g., mean, median, variance or energy) can be estimated as well [44]. Microscopic features can be current harmonics or transient energy [31,45] and require high-sampling frequency to be calculated (1 kHz and above). ...
In this article an energy disaggregation architecture using elastic matching algorithms is presented. The architecture uses a database of reference energy consumption signatures and compares them with incoming energy consumption frames using template matching. In contrast to machine learning-based approaches which require significant amount of data to train a model, elastic matching-based approaches do not have a model training process but perform recognition
using template matching. Five different elastic matching algorithms were evaluated across different
datasets and the experimental results showed that the minimum variance matching algorithm
outperforms all other evaluated matching algorithms. The best performing minimum variance
matching algorithm improved the energy disaggregation accuracy by 2.7% when compared to the
baseline dynamic time warping algorithm.
... We have used real data to simulate some devices such as TV and washing machines. Figure 19.4 presents a 10 s resolution power consumption of a TV [15,25]. At each time step, the Edit command is used to change the parameter of these devices. ...
As the penetration of intelligent and ICT-enabled household devices grows, the need for better understanding of their benefits and threats rises. On the one hand, these devices enable new smart grid applications, such as demand response, which have the potential to improve the usage of energy supply and eventually lead to minimizing the electricity costs. On the other hand, the fine-grained consumption readings can be exploited to reveal private information about the household such as the type of devices and inhabitants behavior. In this paper, we present a co-simulation framework that captures two important worlds of the smart grid, namely the communication world and power world. Real data as well as simulation models are used to simulate several home appliances. The power grid simulator OpenDSS is used to implement the home level power grid, and the data communication simulator OMNeT++ is used to control the behavior of the devices as well as to implement the data communication network. Through a case study, we show how it is possible to integrate privacy approaches inside demand response for a better privacy-preserving smart metering.
... Germany, Austria, Indonesia (N/A) Tracebase (Reinhardt et al., 2012) Dataport (Holcomb, 2012) AMPds (Makonin, Ellert, Bajić, & Popowich, 2016) iAWE (Batra, Gulati, Singh, & Srivastava, 2013) IHEPCDS (Bache & Lichman, 2013) ACS-Fx (Gisler, Ridi, Zufferey, Khaled, & Hennebert, 2013;Ridi, Gisler, & Hennebert, 2014) UK-DALE (Kelly & Knottenbelt, 2015) REFIT (Murray et al., 2015) GREEND ( and energy estimation (EE) algorithms. One possibility is using a resampling technique (e.g., bootstrapping or jackknifing) to generate new labels from the existing ones, that can later be used as training data for event classification. ...
Non‐intrusive load monitoring (also known as NILM or energy disaggregation) is the process of estimating the energy consumption of individual appliances from electric power measurements taken at a limited number of locations in the electric distribution of a building. This approach reduces sensing infrastructure costs by relying on machine learning techniques to monitor electric loads. However, the ability to evaluate and benchmark the proposed approaches across different datasets is key for enabling the generalization of research findings and consequently contributes to the large‐scale adoption of this technology. Still, only recently researchers have focused on creating and standardizing the existing datasets in order to deliver a single interface to run NILM evaluations. Furthermore, there is still no consensus regarding, which performance metrics should be used to measure and report the performance of NILM systems and their underlying algorithms. This paper provides a review of the main datasets, metrics, and tools for evaluating the performance of NILM systems and technologies. Specifically, we review three main topics: (a) publicly available datasets, (b) performance metrics, and (c) frameworks and toolkits. The review suggests future research directions in NILM systems and technologies, including cross‐datasets, performance metrics for evaluation and generalizable frameworks for benchmarking NILM technology.
This article is categorized under:
• Application Areas > Science and Technology
• Application Areas > Data Mining Software Tools
• Technologies > Computational Intelligence
• Technologies > Machine Learning
... ACS-F1 (Appliance Consumption Signatures -Fribourg 1) dataset (Gisler et al., 2013) includes the electrical consumption measurement data (real power (W), reactive power (var), RMS current (A) and phase of voltage relative to current (ϕ)) obtained through two acquisition sessions of one hour on about 100 home appliances. The measurement shave been conducted using plug-based sensors at low frequency (typically every 10 seconds) and the monitored devices include 10 categories of mobile phones, coffee machines, computer stations, fridges and freezers, CD players, lamps, laptops, microwave oven, printers, and televisions. ...
Energy disaggregation, or nonintrusive load monitoring (NILM), aims at estimating the power demand of individual appliances from a household's aggregate electricity consumption. Due to the notable rise in the number of installed smart meters and owing to the numerous advantages of this approach over intrusive methods, NILM has received growing attention in the recent years. In this chapter, after reviewing different categories of household appliances, the state-of-the-art load signatures, including both macroscopic and microscopic features, are introduced. Next, commonly used supervised and unsupervised disaggregation algorithms, which are employed to classify the appliances based on the extracted features, are discussed. Publically accessible datasets and open-source tools, which have been released in the recent years to assist the NILM research and to facilitate the comparison of disaggregation algorithms, are then reviewed. Finally, main applications of energy disaggregation, including providing itemized energy bills, enabling more accurate demand prediction, identifying mal-functioning appliances, and assisting occupancy monitoring, are presented.
... There have been several research studies done on ILM for identification of plug load [24]. Antonio Ridi et.al have developed an appliance consumption signature database of 450 plug load devices from 15 different categories [3,12]. Low frequency sampling at rate of 10 seconds was used to collect data for each device for a period of an hour. ...
Energy utilization can be improved by precise plug load monitoring and control. Plug load energy consumption is nearly 30% of the total building energy consumption. Therefore, plug load identification is a key requirement for energy conservation in buildings. Intrusive load monitoring techniques identify loads precisely but have not been tested widely so far for their performance in changing operating conditions. Hence, the present research proposes a robust low frequency intrusive load monitoring technique to identify load accurately. A smart power strip using proposed load identification technique is designed and developed. Linear regression is applied on the acquired data to capture the behavioral trends of a particular device more explicitly and concisely. Further, weighted K-NN classifier is applied on the transformed data set for device. Experimental results show that the proposed algorithm performs better than the standard classifiers, and can offer tangible savings.
... We based our work on the Appliance Consumption Signature Fribourg 1 (ACS-F1) database [12]. This database contains appliance signatures acquired using low-cost smart plugs capturing the electricity parameters at low frequency with a sampling rate of 10 −1 Hz. ...
We assess the feasibility of unseen appliance recognition through the analysis of their electrical signatures recorded using low-cost smart plugs. By unseen, we stress that our approach focuses on the identification of appliances that are of different brands or models than the one in training phase. We follow a strictly defined protocol in order to provide comparable results to the scientific community. We first evaluate the drop of performance when going from seen to unseen appliances. We then analyze the results of different machine learning algorithms, as the k-Nearest Neighbor (k-NN) and Gaussian Mixture Models (GMMs). Several tunings allow us to achieve 74% correct accuracy using GMMs which is our current best system.
The development of smart grids, traditional power grids, and the integration of internet of things devices have resulted in a wealth of data crucial to advancing energy management and efficiency. Nevertheless, public datasets remain limited due to grid operators' and companies' reluctance to disclose proprietary information. The authors present a comprehensive analysis of more than 50 publicly available datasets, organised into three main categories: micro‐ and macro‐consumption data, detailed in‐home consumption data (often referred to as non‐intrusive load monitoring datasets or building data) and grid data. Furthermore, the study underscores future research priorities, such as advancing synthetic data generation, improving data quality and standardisation, and enhancing big data management in smart grids. The aim of the authors is to enable researchers in the smart and power grid a comprehensive reference point to pick suitable and relevant public datasets to evaluate their proposed methods. The provided analysis highlights the importance of following a systematic and standardised approach in evaluating future methods and directs readers to future potential venues of research in the area of smart grid analytics.
Energy behaviours will play a key role in decarbonising the building sector but require the provision of tailored insights to assist occupants to reduce their energy use. Energy disaggregation has been proposed to provide such information on the appliance level without needing a smart meter plugged in to each load. However, the use of public datasets with pre-collected data employed for energy disaggregation is associated with limitations regarding its compatibility with random households, while gathering data on the ground still requires extensive, and hitherto under-deployed, equipment and time commitments. Going beyond these two approaches, here, we propose a novel data acquisition protocol based on multiplexing appliances’ signals to create an artificial database for energy disaggregation implementations tailored to each household and dedicated to performing under conditions of time and equipment constraints, requiring that only one smart meter be used and for less than a day. In a case study of a Greek household, we train and compare four common algorithms based on the data gathered through this protocol and perform two tests: an out-of-sample test in the artificially multiplexed signal, and an external test to predict the household’s appliances’ operation based on the time series of a real total consumption signal. We find accurate monitoring of the operation and the power consumption level of high-power appliances, while in low-power appliances the operation is still found to be followed accurately but is also associated with some incorrect triggers. These insights attest to the efficacy of the protocol and its ability to produce meaningful tips for changing energy behaviours even under constraints, while in said conditions, we also find that long short-term memory neural networks consistently outperform all other algorithms, with decision trees closely following.
The continual rise in energy consumption has become a significant issue for power systems. Hence it is important to promote energy saving and reduce consumption for solving this problem. Also, finding solutions to enhance energy efficiency is crucial due to the increase in global energy demand. To improve the effectiveness of a power system network, it is important to monitor, manage, and regulate electrical energy needs on a real-time consumption basis. The development of smart homes includes energy management as a key component. A reliable appliance load monitoring (ALM) system is important for the smart energy management framework. ALM's objectives include achieving comprehensive energy sensing and disseminating data on energy use. This will make it possible for automated energy management systems to identify appliances that use higher energy. Thereby enabling them to implement energy saving methods such as rescheduling the high power demand to off-peak hours. Non-intrusive load monitoring (NILM) is one of the most popular energy management strategies in recent years. It enables customers and industries to effectively control their energy use. Utilizing a single energy meter, the NILM technology has gained popularity as an emerging method for tracking events (on/off) and energy use of electrical services and appliances in buildings. Consumers might better understand their appliance behaviour and take the appropriate actions to reduce energy consumption with the aid of information on energy consumption at the appliance level. Hence, the main objective of this paper is to present a comprehensive review of energy management monitoring approaches with an emphasis on non-intrusive load monitoring.
The estimation of electric power utilization, its baseload, and its heating, light, ventilation, and air-conditioning (HVAC) power component, which represents a very large portion of electricity usage in commercial facilities, are important for energy consumption controls and planning. Non-intrusive load monitoring (NILM) is the analytical method used to monitor the energy and disaggregate total electrical usage into appliance-related signals as an alternative to installing multiple electricity meters in the building. However, despite considerable progress, there are a limited number of tools dedicated to the problem of reliable and complete energy disaggregation. This paper presents an experiment consisting in designing an electrical system with electrical energy receivers, and then starting NILM disaggregation using machine learning algorithms (MLA). The quality of this disaggregation was assessed using dedicated indicators. Subsequently, the quality of these MLA was also verified using the available BLUED data source. The results show that the proposed method guarantees non-intrusive load disaggregation but still requires further research and testing. Measurement data have been published as open research data and listed in the literature section repository.
Energy conservation has received a lot of attention in recent ten years. Adoption of sustainable energy is important for meeting energy demand, to address this Non-Intrusive Load Monitoring (NILM) technology is now being developed all over the world. Several energy consumption datasets have been released; each dataset has different properties, uses, and limitations. Thus, a solid comprehension of the relevant datasets will help to improve NILM system. This work is dedicated to detailed study of low frequency (data sampling rate lower than Alternating Current (AC) fundamental frequency) residential datasets, a total eighteen datasets are compared according to their measurement features, collected location, nature of sampling, data collection duration, data development platform etc. Furthermore, datasets are classified depending upon the appliance level and aggregated level data. To full fill the literature gap a new low-cost Indian Energy Dataset with Low frequency (IEDL) has been developed with detailed system deployment. This low frequency dataset collects data from aged appliances (older than ten years), which is allowed to adopt NILM applications in energy saving, recommendation system, appliance behavior, demand prediction area.
Smart nonintrusive load monitoring (NILM) represents a cost‐efficient technology for observing power usage in buildings. It tackles several challenges in transitioning into a more effective, sustainable, and digital energy efficiency environment. This paper presents a comprehensive review of recent trends in the NILM field, in which we propose a multiperspective classification of existing smart NILM techniques. More attention is devoted to describing the contributions of deep learning, feature extraction, computing platforms, and application scenarios for NILM development. Accordingly, NILM technical aspects are first investigated, including data collection devices and public data sets. Next, event‐based and non‐event‐based NILM algorithms are overviewed. Furthermore, potential limitations of existing solutions are identified, highlighting their technical challenges, especially those related to security and privacy preservation, data scarcity, results reproduction, and implementation and business difficulties. Lastly, future directions are explored to overcome the identified limitations.
Early classification of time series has been extensively studied for minimizing class prediction delay in time-sensitive applications such as medical diagnostic and industrial process monitoring. A primary task of an early classification approach is to classify an incomplete time series as soon as possible with some desired level of accuracy. Recent years have witnessed several approaches for early classification of time series. As most approaches have solved the early classification problem using a diverse set of strategies, it becomes very important to make a thorough review of existing solutions. These solutions have demonstrated reasonable performance on a wide range of applications including human activity recognition, gene expression based health diagnostic, and industrial monitoring. In this article, we presenta-systematic review of the current literature on early classification approaches for both univariate and multivariate time series. We divide various existing approaches into four exclusive categories based on their proposed solution strategies. The four categories include prefix based, shapelet based, model based, and miscellaneous approaches. We discuss the applications of early classification and provide a quick summary of the current literature with future research directions. Impact Statement-Early classification is mainly an extension of classification with an ability to classify a time series using limited data points. It is true that one can achieve better accuracy if one waits for more data points, but opportunities for early interventions could equally be missed. In a pandemic situation such as COVID-19, early detection of an infected person becomes more desirable to curb the spread of the virus and possibly save lives. Early classification of gas (e.g., methyl isocyanate) leakage can help to avoid life-threatening consequences on human beings. Early classification techniques have been successfully applied to solve many time-critical problems related to medical diagnostic and industrial monitoring. This article provides a systematic review of the current literature on these early classification approaches for time series data, along with their potential applications. It also suggests some promising directions for further work in this area.
A third of the final electricity in the EU is consumed by households. The increased usage of multiple electrical devices, electromobility, self-generation and consumption of electricity as well as work-from-home will fundamentally change the residential electricity load profile, so a deep understanding of the current state of residential electricity load profile modelling is necessary. The objective of this paper is to perform a literature review, evaluate the current state of the residential electricity load profile modelling, categorise the models, propose future research directions and applications, identify the challenges the researchers face when building these models and offer possible solutions. Thirty two residential electricity load profile models are identified and a new definition of the residential electricity load profile model is proposed. A new categorisation system based on the identification of the main features of these thirty two studies is introduced. Future research directions and applications are presented and the most important challenges that modellers face when attempting to build such models are identified and discussed. The most important challenge identified is the privacy concerns of the participants or potential participants. These concerns are at least partially responsible for the existence of the rest of the challenges. The creation and implementation of an anonymisation algorithm, before any human has access to any measured datasets, the implementation of a crowd sourcing approach which addresses the privacy concerns of the citizens and increased funding for the installation of privacy-proof smart-meters by the public and measurement campaigns are identified as possible solutions to the challenges faced by modellers.
Building energy consumption accounts for a large fraction of the total global energy usage, and considerable energy savings are expected to be achieved in this respect through residential electrical load monitoring. Due to the limitations on the practical implementation of in-depth and expensive monitoring systems, non-intrusive load monitoring (NILM) is becoming a hot topic. In this paper, an overview of the state of the art residential electrical load monitoring is presented. Different from previous reviews, the applications of load monitoring are particularly addressed, based on which, technical challenges of load monitoring techniques, including NILM, are identified and thoroughly discussed, together with possible developments and trends predicted from the authors’ perspective.
Early classification of time series has been extensively studied for minimizing class prediction delay in time-sensitive applications such as healthcare and finance. A primary task of an early classification approach is to classify an incomplete time series as soon as possible with some desired level of accuracy. Recent years have witnessed several approaches for early classification of time series. As most of the approaches have solved the early classification problem with different aspects, it becomes very important to make a thorough review of the existing solutions to know the current status of the area. These solutions have demonstrated reasonable performance in a wide range of applications including human activity recognition, gene expression based health diagnostic, industrial monitoring, and so on. In this paper, we present a systematic review of current literature on early classification approaches for both univariate and multivariate time series. We divide various existing approaches into four exclusive categories based on their proposed solution strategies. The four categories include prefix based, shapelet based, model based, and miscellaneous approaches. The authors also discuss the applications of early classification in many areas including industrial monitoring, intelligent transportation, and medical. Finally, we provide a quick summary of the current literature with future research directions.
Smart meter roll-outs provide easy access to granular meter measurements, enabling advanced energy services, ranging from demand response measures, tailored energy feedback and smart home/building automation. To design such services, train and validate models, access to data that resembles what is expected of smart meters, collected in a real-world setting, is necessary. The REFIT electrical load measurements dataset described in this paper includes whole house aggregate loads and nine individual appliance measurements at 8-second intervals per house, collected continuously over a period of two years from 20 houses. During monitoring, the occupants were conducting their usual routines. At the time of publishing, the dataset has the largest number of houses monitored in the United Kingdom at less than 1-minute intervals over a period greater than one year. The dataset comprises 1,194,958,790 readings, that represent over 250,000 monitored appliance uses. The data is accessible in an easy-to-use comma-separated format, is time-stamped and cleaned to remove invalid measurements, correctly label appliance data and fill in small gaps of missing data.
In recent years, as Cloud Computing and Internet of Things (IoT) technologies develop rapidly, smart home technology has entered a new stage: smart electric power usage, which will break the traditional extensive management situation for electric power loads, particularly under the background of an energy crisis. People will gradually gravitate towards the demands for more energy-related and intelligent services in smart home systems such as energy saving, improved understanding of electrical consumption and electrical safety of appliances. When the power consumption is considered by a smart grid, detailed electrical information of power loads will be perceived for demand management and optimization. To address these issues, measuring and monitoring the power of residential individual appliances are highly demanded by many applications.
The goal of the Smart* project is to optimize home energy con-sumption. As part of the project, we have designed and deployed a "live" system that continuously gathers a wide variety of envi-ronmental and operational data in three real homes. In contrast to prior work, our focus has been on sensing depth, i.e., collecting as much data as possible from each home, rather than breadth, i.e., collecting data from as many homes as possible. Our data captures many important aspects of the home environment, including aver-age household electricity usage every second, as well as usage at every circuit and nearly every plug load, electricity generation data from on-site solar panels and wind turbines, outdoor weather data, temperature and humidity data in indoor rooms, and, finally, data for a range of important binary events, e.g., at wall switches, the HVAC system, doors, and from motion sensors. We also have elec-tricity usage data every minute from 400 anonymous homes. This data corpus has served as the foundation for much of our recent research. In this paper, we describe our data sets as well as basic software tools we have developed to facilitate their collection. We are releasing both the data and tools publicly to the research com-munity to foster future research on designing sustainable homes.
Dynamic load management, i.e., allowing electricity utilities to remotely turn electric appliances in households on or off, represents a key element of the smart grid. Appliances should however only be disconnected from mains when no negative side effects, e.g., loss of data or thawing food, are incurred thereby. This motivates the use of appliance identification techniques, which determine the type of an attached appliance based on the continuous sampling of its power consumption. While various implementations based on different sampling resolutions have been presented in existing literature, the achievable classification accuracies have rarely been analyzed. We address this shortcoming and evaluate the accuracy of appliance identification based on the characteristic features of traces collected during the 24 hours of a day. We evaluate our algorithm using more than 1,000 traces of different electrical appliances' power consumptions. The results show that our approach can identify most of the appliances at high accuracy.
We report on the development of an innovative system which can
automatically recognize home appliances based on their electric consumption
profiles. The purpose of our system is to apply adequate rules to control
electric appliance in order to save energy and money. The novelty of our
approach is in the use of plug-based low-end sensors that measure the
electric consumption at low frequency, typically every 10 seconds. Another
novelty is the use of machine learning approaches to perform the
classification of the appliances. In this paper, we present the system
architecture, the data acquisition protocol and the evaluation framework.
More details are also given on the feature extraction and classification
models being used. The evaluation showed promising results with a correct
rate of identification of 85\%.
Evaluating speaker verification algorithms on relevant speech corpora is a key issue for measuring the progress and discovering the remaining difficulties of speaker verification systems. A common evaluation framework is also a key point when comparing systems produced by different labs. The speech group of the National Institute of Standards and Technology (NIST) has been organizing evaluations of text-independent telephony speaker verification technologies since 1997, with an increasing success and number of participants over the years. These NIST evaluations have been recognized by the speaker verification scientific community as a key factor for the improvement of the algorithms over the last decade. However, these evaluations measure exclusively the effectiveness in term of performance of the systems, assuming some conditions of use that are sometimes far away from any real-life commercial context for telephony applications. Other important aspects of speaker verification systems are also ignored by such evaluations, such as the efficiency, the usability and the robustness of the systems against impostor attacks. In this paper we present a review of the current NIST speaker verification evaluation methods, trying to put objectively into evidence their current benefits and limitations. We also propose some concrete solutions for going beyond these limitations.
Activity sensing in the home has a variety of important applications, including healthcare, entertainment, home automation, energy monitoring and post-occupancy research studies. Many existing systems for detecting occupant activity require large numbers of sensors, invasive vision systems, or extensive installation procedures. We present an approach that uses a single plug-in sensor to detect a variety of electrical events throughout the home. This sensor detects the electrical noise on residential power lines created by the abrupt switching of electrical devices and the noise created by certain devices while in operation. We use machine learning techniques to recognize electrically noisy events such as turning on or off a particular light switch, a television set, or an electric stove. We tested our system in one home for several weeks and in five homes for one week each to evaluate the system performance over time and in different types of houses. Results indicate that we can learn and classify various electrical events with accuracies ranging from 85-90%.
In automated energy management systems, to make instantaneous decisions based on the appliance status information, continuous data access is a key requirement. With the advances in sensor and communication technologies, it is now possible to remotely monitor the power consumption data. However, before an appliance is actively monitored, it must be identified using the obtained power consumption data. Appropriate methods are required to analyse power consumption patterns for proper appliance recognition. The focus of this work is to provide the model structure for storing and distinguishing the recurring footprints of the household appliances. Hidden Markov model based method is proposed to recognize the individual appliances from combined load. It is found that the proposed method can efficiently differentiate the power consumption patterns of appliances from their combined profiles.
Microgrids are well-suited for electrification of remote off-grid areas. This paper sketches the concept of a plug-and-play microgrid with a minimum of configuration effort needed for setup. When the load of such an off-grid microgrid grows over the generation capacity and energy storage is not sufficient, demand has to be reduced to prevent a blackout. In order to decide which loads are inessential and can be shedded, automated load recognition on the basis of measured power consumption profiles is needed. Two promising approaches from the area of speech recognition, Dynamic Time Warping and Hidden Markov Models, are compared for this application. It is found that a key feature to achieve good recognition efficiency is a careful selection of the features extracted from the measured power data.
Activity sensing in the home has a variety of important applications, including healthcare, entertainment, home automation, energy monitoring and post-occupancy research studies. Many existing systems for detecting occupant activity require large numbers of sensors, invasive vision systems, or extensive installation procedures. We present an approach that uses a single plug-in sensor to detect a variety of electrical events throughout the home. This sensor detects the electrical noise on residential power lines created by the abrupt switching of electrical devices and the noise created by certain devices while in operation. We use machine learning techniques to recognize electrically noisy events such as turning on or off a particular light switch, a television set, or an electric stove. We tested our system in one home for several weeks and in five homes for one week each to evaluate the system performance over time and in different types of houses. Results indicate that we can learn and classify various electrical events with accuracies ranging from 85-90%.
This paper considers additive factorial hidden Markov models, an extension to HMMs where the state factors into multiple independent chains, and the output is an additive function of all the hidden states. Although such models are very powerful, accurate inference is unfortunately difficult: exact inference is not computationally tractable, and existing approximate inference techniques are highly susceptible to local optima. In this paper we propose an alternative inference method for such models, which exploits their additive structure by 1) looking at the observed difference signal of the observation, 2) incorporating a “robust” mixture component that can account for unmodeled observations, and 3) constraining the posterior to allow at most one hidden state to change at a time. Combining these elements we develop a convex formulation of approximate inference that is computationally efficient, has no issues of local optima, and which performs much better than existing approaches in practice. The method is motivated by the problem of energy disaggregation, the task of taking a whole home electricity signal and decomposing it into its component appliances; applied to this task, our algorithm achieves state-of-the-art performance, and is able to separate many appliances almost perfectly using just the total aggregate signal.
Energy and sustainability issues raise a large number of problems that can be tackled using approaches from data mining and machine learning, but traction of such problems has been slow due to the lack of publicly available data. In this paper we present the Reference Energy Disaggregation Data Set (REDD), a freely available data set containing de-tailed power usage information from several homes, which is aimed at furthering research on energy disaggregation (the task of determining the component appliance contributions from an aggregated electricity signal). We discuss past ap-proaches to disaggregation and how they have influenced our design choices in collecting data, we describe the hardware and software setups for the data collection, and we present initial benchmark disaggregation results using a well-known Factorial Hidden Markov Model (FHMM) technique.
Today's solutions to inform residents about their electricity consumption are mostly confined to displaying aggregate readings collected at meter level. A reliable identification of appliances that require disproportionate amounts of energy for their operation is generally unsupported by these systems, or at least requires significant manual configuration efforts. We address this challenge by placing low-cost measurement and actuation units into the mains connection of appliances. The distributed sensors capture the current flow of individual appliances at a sampling rate of 1.6kHz and apply local signal processing to the readings in order to extract characteristic fingerprints. These fingerprints are communicated wirelessly to the evaluation server, thus keeping the required airtime and energy demand of the transmission low. The evaluation server employs machine learning techniques and caters for the actual classification of attached electric appliances based on their fingerprints, enabling the correlation of consumption data and the appliance identity. Our evaluation is based on more than 3,000 current consumption fingerprints, which we have captured for a range of household appliances. The results indicate that a high accuracy is achieved when locally extracted current consumption fingerprints are used to classify appliances.
Home energy displays are emerging home energy management devices. Their energy saving potential is limited, because most display whole-home electricity consumption data. We propose a new approach to disaggregation electricity consumption by individual appliances and/or end uses that would enhance the effectiveness of home energy displays. The proposed method decomposes a system of appliance models into tuplets of appliances overlapping in power draw. Each tuplet is disaggregated using a modified Viterbi algorithm. In this way, the complexity of the disaggregation algorithm is linearly proportional to the number of appliances. The superior accuracy of the method is illustrated by a simulation example and by actual household data.
Home energy displays are emerging home energy management devices. However, their energy savings potential is limited, because most display whole-home electricity consumption data. We propose a new approach to disaggregation electricity consumption by individual appliances and/or end uses that would enhance the effectiveness of home energy displays.
Load signatures embedded in common electricity consumption patterns, in fact, could render much information pertaining to the nature of the appliances and their usage patterns. Based on the proposed disaggregation framework, we use three advanced disaggregation algorithms, called committee decision mechanisms (CDMs), to perform load disaggregation at the metering level. Three random switching simulators are also developed to investigate the performance of different CDMs under a variety of scenarios. Through Monte Carlo simulations, we demonstrate that all CDMs outperform any single-feature, single-algorithm-based disaggregation methods. With sensitivity analysis, we also show that the CDMs are less sensitive to any load dynamics and noise. We finally demonstrate some applications of this technology in terms of appliance usage tacking and estimated energy consumption of each appliance.
The paper illustrates an artificial neural network (ANN) based procedure for the identification of pattern-of-use of some main domestic electric appliances from daily profiles of energy recordings taken at household's meter panel at 15 minute steps. The paper describes the architecture used which is structured into multiple subsequent stages based on ANNs. The application of the procedure to some real daily load diagrams as recorded at few household meters demonstrates the efficiency of the proposed approach. The utility of ANNs for extracting further useful information concerning energy usage from data bases typically available for electric companies is finally discussed.
This tutorial paper describes the fundamental mechanism by which mechanical switches produce electrical "noise," and the parameters which determine the complex waveforms, frequency componets, and amplitudes produced. Understanding the fundamentals removes some of the mystery often associated with this ubiquitous source of noise and transients, and can assist in development of effective methods of preventing resultant malfunctions and damage in electronic equipment. The primary focus is upon switches operating low-current inductive loads on 120-V residential power lines, and shows peak amplitudes up to several kilovolts and frequencies up to several hundred megahertz produced on the supply line.
A nonintrusive appliance load monitor that determines the energy
consumption of individual appliances turning on and off in an electric
load, based on detailed analysis of the current and voltage of the total
load, as measured at the interface to the power source is described. The
theory and current practice of nonintrusive appliance load monitoring
are discussed, including goals, applications, load models, appliance
signatures, algorithms, prototypes field-test results, current research
directions, and the advantages and disadvantages of this approach
relative to intrusive monitoring
- G David
- O Stork Richard
- Peter E Duda
- Hart
David G. Stork Richard O. Duda, Peter E. Hart,
Pattern Classification (2nd Edition),
Wiley-
Interscience, 2000.
Are you Energy-Efficient: Sense it on theWeb!
- D Guinard
- M Weiss
- V Trifa
Residential electricity use feedback: A research synthesis and economic framework
- B Neenan
Plogg wireless energy management
- Energy Optimizers Limited
Energy Optimizers Limited, "Plogg wireless energy
management," http://www.plogg.co.uk.
Machine learning approaches for electric appliance classification
- D Zufferey
- C Gisler
- O Abou Khaled
- J Hennebert
D. Zufferey, C. Gisler, O. Abou Khaled, and J. Hennebert, "Machine learning approaches for electric appliance classification," in Proceedings of
the 11th International Conference on Information
Science, Signal Processing and their Applications
(ISSPA'12), Montreal, Canada, July 2012, pp. 740-745.
Blued: A fully labeled public dataset for event-based non-intrusive load monitoring research
- A Kyle
- A Filip
- D Bentez
- D Carlson
- A Rowe
- M Bergs
A. Kyle, A. Filip, D. Bentez, D. Carlson, A. Rowe,
and M. Bergs, "Blued: A fully labeled public
dataset for event-based non-intrusive load monitoring research," in Proceedings of the ACM Workshop on Data Mining Applications in Sustainability
(SustKDD 2012), Beijing, China, Aug. 2012.
Smart*: An open data set and tools for enabling research in sustainable homes
- S Barker
- A Mishra
- D Irwin
- E Cecchet
- P Shenoy
S. Barker, A. Mishra, D. Irwin, E. Cecchet, and
P. Shenoy, "Smart*: An open data set and tools for
enabling research in sustainable homes," in Proceedings of the ACM Workshop on Data Mining Applications in Sustainability (SustKDD 2012), Beijing,
China, Aug. 2012.