ArticlePDF Available

A Practical Approach to Residential Appliances On-line Anomaly Detection: A Case Study of Standard and Smart Refrigerators

Authors:

Abstract and Figures

Anomaly detection is a significant application of residential appliances load monitoring systems. As an essential prerequisite of load diagnosis services, anomaly detection is critical to energy saving and occupant comfort actualization. Notwithstanding, the investigation into diagnosis of household anomalous appliances has not been decently taken into consideration. This paper presents an extensive study about operation-time anomaly detection of household devices particularly, refrigerators, in terms of appliances candidate, by utilizing their energy consumption data. Energy as a quantitative property of electrical loads, is a reliable information for a robust diagnosis. Additionally, it is very practical since it is low-priced to measure and definite to interpret. Subsequently, an on-line anomaly detection approach is proposed to effectively determine the anomalous operation of the household appliances candidate. The proposed approach is capable of continuously monitoring energy consumption and providing dynamic information for anomaly detection algorithms. A machine learning-based technique is employed to construct efficient models of appliances normal behavior with application to operation-time anomaly detection. The performance of the suggested approach is evaluated through a set of diagnostic tests, by utilizing normal and anomalous data of targeted devices, measured by an acquisition system. In addition, a comparison analysis is provided in order to further examine the effectiveness of the developed mechanism by exploiting a public database. Moreover, this study elaborates sensible remarks on an effective management of anomaly detection and diagnosis decision phases, pivotal to correctly recognition of a faulty/abnormal operation. Indeed, through experimental results of case studies, this work assists in the development of a load monitoring and anomaly detection system with practical implementation.
Content may be subject to copyright.
A preview of the PDF is not available
... To find anomalous power consumption, the authors trained a deep neural network to recognize the extracted micro-moments classes, including the anomalous ones. Hosseini et al. in [14] introduced a semi-supervised framework for detecting electrical anomalies in the power consumption of a fridge. The authors characterized the fridge's power profile in terms of its daily energy consumption. ...
... Himeur et al. [13] Supervised Adaptive Manual Hosseini et al. [14] Unsupervised Adaptive Manual Benninger et al. [15] Unsupervised hard-coded Manual Rashid et al. [16] Unsupervised Adaptive Manual Proposed solution Unsupervised Adaptive Automatic on the study of anomalous appliances by extending the set of analyzed devices to more energy-intensive equipments. In particular, instead of studying the hourly energy consumption of the monitored devices, we want to investigate the characteristics of their instantaneous power measurements, which provide a greater level of detail. ...
Article
Electrical anomalies in residential buildings represent a serious problem that can unpredictably change the power profiles of end-users, causing a sub-optimal energy distribution. In addition, electrical faults can cause unnoticed energy wastages and higher energy bills, or even severe damages for properties and people in the most critical cases. In this paper, we introduce a novel anomaly detection method for detecting electrical faults in household appliances based on the analysis of their power signatures with unsupervised deep learning techniques. For this purpose, we trained a variational autoencoder to reconstruct the power signatures of three commonly used devices: the dishwasher, the washing machine and the dryer. For each use case, we injected several randomly generated anomalies that simulate to our best realistic electrical faults in these devices. To demonstrate the effectiveness of our method, we compared the accuracy of the variational autoencoder with the classification performance of a one-class support vector machine (OC-SVM) trained with two manual features: the energy consumption and duration of the appliance’s operations. The variational autoencoder showed higher classification accuracy with respect to the OC-SVM, reporting an F1-score greater than 90% in all the use cases. Most importantly, the results demonstrate that deep anomaly detection methods outperform traditional algorithms based on handcrafted features, allowing to better characterize the set of normal cycles and produce more precise alerts for the monitored devices.
... The use of predictions for control has been shown to provide more optimum results than conventional controls, for example by using a prediction model based on the Monte Carlo method, radial basis function neural network (RBF-NN), and Markovchain [10]. In addition to the use of predictions for control, the detection of anomalies at refrigerator temperature is also another area of research that is considered [11]. One of the benefits of anomaly detection (AD) in refrigerator monitoring is to ensure that the stored supply is maintained safely [12]. ...
... In addition to focusing on the IoT architecture and monitoring the contents of the refrigerator, other studies also discuss the detection of anomalies in refrigerator temperatures, such as in paper [11], which applies an on-line based AD. However, in this study, the motivating urgency to detect anomalies is the detection of unusual loads. ...
... Hosseini at al. in [13] propose a semi-supervised anomaly detection framework for detecting faulty behaviours in the refrigerators. Their algorithm characterises the fridge by means of its daily energy consumption and average power consumption. ...
... Furthermore, the supervised approaches have been always evaluated by using artificially injected anomalies, which can hardly depict an authentic anomalous behaviour. On the other hand, the unsupervised solution presented in [13] is quite promising from a feasibility point of view and it was tested on real fridges with true anomalies. However, it would be interesting to test the unsupervised approach on additional appliances. ...
... In the first phase, the most probable behavior of a thermostat is computed by analyzing its patterns. Accordingly, the Kernel Density Estimation (KDE) method is applied to thermostat setpoints to select the temperature profile with the highest probability [18]. This profile presents the setpoint temperature candidate in its associated thermal zone. ...
Conference Paper
Full-text available
This paper examines the impact of building zoning on demand flexibility potentials. The building thermal dynamic response is modeled by using the state-space representation method. The comfort preference of the occupants is defined by statistical analysis of set-point temperature patterns. The created models are used to study energy flexibility through a coordinated energy management problem by means of the Stackelberg game strategy. The zoning influence on flexibility potentials is analyzed by applying this procedure to a set of houses in Quebec based on two specific scenarios accounting for single and multi-zones. The flexibility potential is evaluated by comparing overall demand peak reduction and heating consumption under each scenario signifying passive heating storage capacity of buildings. The single-zone scenario performs better at harvesting energy flexibility through reducing the peak load by 33%. The results demonstrate that the building zoning definition can lead to conditions under which the demand flexibility can change.
... Wang et al. [25] proposed a semi-supervised learning anomaly detection model, sample efficient home power anomaly detection (SEPAD), in which the k-means and z-score function [26] were used to point out the suspicious data, and a semi-SVM based pattern matching algorithm was proposed to identify anomaly power consumption events. Hosseini et al. [27] focused on the appliance-level anomaly detection and trained the classification modes by using the operation patterns for the refrigerators depending on the semi-supervised learning strategy. Fan et al. [28] proposed a building electricity anomaly detection model based on unsupervised classification to reduce the training cost lower than that of supervised learning-based models. ...
Article
Full-text available
Due to the climate crisis, energy-saving issues and carbon reduction have become the top priority for all countries. Owing to the increasing popularity of advanced metering infrastructure and smart meters, the cost of acquiring data on residential electricity consumption has substantially dropped. This change promotes the analysis of residential electricity consumption, which features both small and complicated consumption behaviors, using machine learning to become an important research topic among various energy saving and carbon reduction measures. The main subtopic of this subject is the identification of abnormal electricity consumption behaviors. At present, anomaly detection is typically realized using models based on low-level features directly collected by sensors and electricity meters. However, due to the significant number of dimensions and a large amount of redundant information in these low-level features, the training efficiency of the model is often low. To overcome this, this study adopts an autoencoder, which is a deep learning technology, to extract the high-level electricity consumption information of residential users to improve the anomaly detection performance of the model. Subsequently, this study trains one-class SVM models for anomaly detection by using the high-level features of five actual residential users to verify the benefits of high-level features.
... Besides the monitored electrical quantities, for some application areas, the existence of labeled appliance transitions (also referred to as power events) is essential to train and validate the underlying algorithms. This is the case of real-time appliance recognition algorithms that rely on the accurate detection and classification of appliance transitions 20,21 , and anomaly detection algorithms that often rely on historical patterns of appliance transitions 22,23 . Still, to the best of our knowledge, to date, there are only four real-world datasets that contain labeled appliance transitions, namely, Building-Level Fully-labeled dataset for Electricity Disaggregation (2022) 9:134 | https://doi.org/10.1038/s41597-022-01252-2 ...
Article
Full-text available
Smart meter data is a cornerstone for the realization of next-generation electrical power grids by enabling the creation of novel energy data-based services like providing recommendations on how to save energy or predictive maintenance of electric appliances. Most of these services are developed on top of advanced machine-learning algorithms, which rely heavily on datasets for training, testing, and validation purposes. A limitation of most existing datasets, however, is the scarcity of labels. The SustDataED2 dataset described in this paper contains 96 days of aggregated and individual appliance consumption from one household in Portugal. The current and voltage waveforms were sampled at 12.8 kHz, and the individual consumption of 18 appliances was sampled at 0.5 Hz. The dataset also contains the timestamps of the ON-OFF transitions of the monitored appliances for the entire deployment duration, providing the necessary ground truth for the evaluation of machine learning problems, particularly Non-Intrusive Load Monitoring. The data is accessible in easy-to-use audio and comma-separated formats.
... To this end, computers or applications are utilized in order to manage the functions and features of a main appliance remotely via the Internet [11]. At a smart home, the kitchen is the most likely place to house smart devices [12], and the refrigerator is one of the most important appliances in the kitchen [13][14][15]. ...
Article
Full-text available
With the growing need of technology into varied fields, dependency is getting directly proportional to ease of user-friendly smart systems. The advent of artificial intelligence in these smart systems has made our lives easier. Several Internet of Things- (IoT-) based smart refrigerator systems are emerging which support self-monitoring of contents, but the systems lack to achieve the optimized run time and data security. Therefore, in this research, a novel design is implemented with the hardware level of integration of equipment with a more sophisticated software design. It was attempted to design a new smart refrigerator system, which has the capability of automatic self-checking and self-purchasing, by integrating smart mobile device applications and IoT technology with minimal human intervention carried through Blynk application on a mobile phone. The proposed system automatically makes periodic checks and then waits for the owner’s decision to either allow the system to repurchase these products via Ethernet or reject the purchase option. The paper also discussed the machine level integration with artificial intelligence by considering several features and implemented state-of-the-art machine learning classifiers to give automatic decisions. The blockchain technology is cohesively combined to store and propagate data for the sake of data security and privacy concerns. In combination with IoT devices, machine learning, and blockchain technology, the proposed model of the paper can provide a more comprehensive and valuable feedback-driven system. The experiments have been performed and evaluated using several information retrieval metrics using visualization tools. Therefore, our proposed intelligent system will save effort, time, and money which helps us to have an easier, faster, and healthier lifestyle.
Article
Worldwide, buildings consume a large amount of energy, and a significant share of this energy is wasted due to system degradation, inappropriate control systems and improper maintenance activities. Fortunately, buildings have become data-intensive, which has led to utilising operation data with the help of data mining technologies to improve building energy performance and cut down on energy waste in the operational stage. A systematic review was undertaken as per the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) procedure to identify studies reporting the relationship between building energy performance and maintenance strategies. The search was applied to five search engines: Compendex, Inspec, Geobase, GeoRef and Web of Science. A total of 192 relevant articles were selected, namely 134 journal articles, 44 conference papers, and 14 review papers. This work highlighted the building energy performance deficiencies, the different condition monitoring techniques (top-down and bottom-up), and the types of faults. Also, the types of input data, algorithms, and the evaluation methods of the proposed strategies were investigated. The operation data is the most popular data form used in the case studies, especially with model-based and data-based techniques, and chiller plants were the most studied systems. Finally, the bibliometric method was used to identify the most influential articles within the field, which could significantly aid the planning of future research activities and networking.
Article
Full-text available
In the current energy ecosystem, the need for a Hybrid Appliance Load Monitoring System (HALMS) to establish a smarter grid and energy infrastructure is undeniable. The increasing popularity of the Internet of Things (IoT) has suddenly pushed the demand for smart and connected devices. This review introduces the term smart plug as a device that uses IoT for establishing HALMS. These smart plugs are a handy solution to make the so-called 'dumb' devices smart. The strategy of smart plugs to enhance the energy management experience in connected spaces is presented. This study extensively highlights the current smart plug technologies and the relevant activities and limitations that need to overcome the requirements of HALMS.
Article
Full-text available
This study seeks to understand the thermal and energetic behavior of a domestic refrigerator more widely by experimentally evaluating the main effects of the thermal load (food) and the variation of the ambient temperature. To carry out the experiments, the thermal load was classified based on the results of a survey conducted on different consumers in the state of Guanajuato, Mexico. The thermal behavior of both compartments of the refrigerator, the total energy consumption, the power of the compressor in its first on-state, and the coefficient of performance, according to the classification of the thermal loads and the room temperature, were evaluated. Finally, it is verified that the thermal load and the room temperature have a significant influence on the energy performance of the refrigerator.
Article
Full-text available
Faults in buildings systems affect energy efficiency and occupancy comfort. Simulating building behavior and comparing it with measured data allows to detect discrepancies due to faults. We propose a methodology to recursively compare actual data with dynamic energy simulations at different layers of aggregation to reduce the scope in searching for faults through the development the Online Energy Simulator, a tool to set up automated simulations using standard interfaces usable with different building systems and simulation engines. We test our simulator on a real building at the University of Southern Denmark, showing how continuous monitoring allows to quickly detect and identify buildings faults.
Article
Enabling diagnosis capabilities of Appliance Load Monitoring (ALM) necessitates providing in-operation information of appliances’ behavior. Due to both appliances’ time-varying model parameters and operations, household aggregated consumption has a dynamic structure. Existing time-invariant load models, built of off-line datasets with static information, are not sufficient to capture the actual behavior of the power consumption. In fact, these models, generally obtained from exhaustive training phases are intended to satisfy load monitoring goals. Therefore, a time-variant load modeling is more practical to capture such a dynamic property of the power consumption. Accordingly, this paper presents an adaptive on-line appliance-level load modeling approach, to design a load monitoring structure for diagnosis purposes. By using the aggregated power consumption of individual households, our proposed structure results in an autonomous household database construction. The modeling procedure begins with a designed recurrent pattern recognition system that is capable of detecting and maintaining load models. This load model structure is determined by using a hidden Markov model (HMM) with dynamic parameters, that are extracted from aggregated signal and trained within an on-line learning process. Our proposed approach can detect time-varying power consumption behavior and estimate the robust load models of appliances. Additionally, our novelty in employing a set of straightforward algorithms, suggests the practicality of our database construction approach.
Article
Identification of faulty appliance behaviour in real time can signal energy wastage and the need for appliance servicing or replacement leading to energy savings. The problem of appliance fault or anomaly detection has been tackled vastly in relation to submetering, which is not scalable since it requires separate meters for each appliance. At the same time, for applications such as energy feedback, Non-intrusive load monitoring (NILM) has been recognised as a scalable and practical alternative to submetering. However, the usability of NILM for anomaly detection has not yet been investigated. Since the goal of NILM is to provide energy consumption estimate, it is unclear if the signal fidelity of appliance signatures generated by state-of-the-art NILM is sufficient to enable accurate appliance fault detection. In this paper, we attempt to determine whether appliance signatures detected by NILM can be used directly for anomaly detection. This is carried out by proposing an anomaly detection algorithm which performs well for submetering data and evaluate its ability to identify the same faulty behaviour of appliances but with NILM-generated appliance power traces. Our results on a dataset of six residential homes using four state-of-the-art NILM algorithms show that, on average, NILM traces are not as robust to identification of faulty behaviour as compared to using submetered data. We discuss in detail observations pertaining to the reconstructed appliance signatures following NILM and their fidelity with respect to noise-free submetered data.
Conference Paper
Buildings across the world contribute about one-third of the total energy consumption. Studies report that anomalies in energy consumption caused by faults and abnormal appliance usage waste up to 20% of energy in buildings. Recent works leverage smart meter data to find such anomalies; however, such works do not identify the appliance causing the anomaly. Moreover, most of these works are not real-time and report the anomaly at the end of the day. In this paper, we propose a technique named Rimor that addresses these limitations. Rimor predicts the energy consumption of a home using historical energy data and contextual information and flags an anomaly when the actual energy consumption deviates significantly from the predicted consumption. Further, it identifies anomalous appliance(s) by using easy-to-collect appliance power ratings. We evaluated it on four real-world energy datasets containing 51 homes and found it to be 15% more accurate in detecting anomalies as compared to four other baseline approaches. Rimor reports an appliance identification accuracy of 82%. In addition, we also release an anomaly annotated energy dataset for the research community.
Article
Changes in system topology, such as branch breaking and the loss of a generator or load, may profoundly influence the operation security of the power system. This study introduces a novel deep-learning based fault diagnosis method using power flow to diagnose topology changes in the power system. Power flow samples with different system states and topologies are first computed numerically; then, they are transformed into computer-visualised images. Using massive power-flow image samples, a convolutional neural network that aims to identify the system state is trained. A feature-map restriction technique is used to restructure the network. To enhance the robustness of the network, the random noise of branch flow is considered in the sample generation process. The results show that the proposed deep-learning based method may diagnose system faults effectively.
Article
Many hyperspectral anomaly detectors are designed based on the traditional Mahalanobis distance-based RX algorithm, which is usually considered as an inverse operation of the principal components analysis. Such detectors include the uniform target detector (UTD) algorithm, RX-UTD algorithm, and so on. However, the possibility of background statistical contamination caused by anomalies still exists. In order to alleviate this problem, in this paper, we propose a spectral matched filter (background-anomaly component projection and separation optimized filter) to minimize the average output energy of separate image components and the output values of the weighted background regular term for hyperspectral image anomaly detection, which could strengthen the separation between anomalies and backgrounds. By calculating the optimal solution to the background-anomaly component projection and separation function, we obtain the optimal projection, where we can effectively suppress the background while highlighting the anomalies. Proposed algorithm has the following research advantages: 1) it creates a novel collaborative component projection and robust background optimization function to separate the background and anomalies and 2) it analyzes the intrinsic statistical distribution of pixels and applies appropriate iterative shrinkage-thresholding algorithm to solve the ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -min problem. Experiments were conducted on three real hyperspectral data sets. The detection results demonstrate that the proposed algorithm is superior to other state-of-the-art anomaly detection algorithms.