Non-intrusive Load Monitoring Approach Research Articles

Review on Deep Neural Networks Applied to Low-Frequency NILM

This paper reviews non-intrusive load monitoring (NILM) approaches that employ deep neural networks to disaggregate appliances from low frequency data, i.e., data with sampling rates lower than the AC base frequency. The overall purpose of this review is, firstly, to gain an overview on the state of the research up to November 2020, and secondly, to identify worthwhile open research topics. Accordingly, we first review the many degrees of freedom of these approaches, what has already been done in the literature, and compile the main characteristics of the reviewed publications in an extensive overview table. The second part of the paper discusses selected aspects of the literature and corresponding research gaps. In particular, we do a performance comparison with respect to reported mean absolute error (MAE) and F1-scores and observe different recurring elements in the best performing approaches, namely data sampling intervals below 10 s, a large field of view, the usage of generative adversarial network (GAN) losses, multi-task learning, and post-processing. Subsequently, multiple input features, multi-task learning, and related research gaps are discussed, the need for comparative studies is highlighted, and finally, missing elements for a successful deployment of NILM approaches based on deep neural networks are pointed out. We conclude the review with an outlook on possible future scenarios.

A non-intrusive load monitoring approach for very short-term power predictions in commercial buildings

In this study, a new algorithm is developed to extract device profiles in a fully unsupervised manner from three-phases reactive and active aggregate power measurements. The extracted device profiles are then applied to disaggregate the aggregate power measurements by means of particle swarm optimization. Then, a new approach to very short-term power predictions is presented, which makes use of the disaggregation data. For this purpose, a state change forecast is carried out for each device by an artificial neural network and subsequently converted into a power prediction by reconstructing the power profile with respect to the state changes and device profiles. The forecast horizon is 15 min. In order to demonstrate the developed approaches, three-phase reactive and active aggregate power measurements of a multi-tenant commercial building are employed as a case study. The granularity of the data used is 1 s. In total, 52 device profiles are extracted from the aggregate power data. The disaggregation exhibited a highly accurate reconstruction of the measured power with an energy percentage error of approximately 1 %. The indirect power prediction method developed is then applied to the measured power data and outperforms the two persistence forecasts, as well as an artificial neural network designed for 24h-ahead power predictions working in the power domain.

Non-invasive load-shed authentication model for demand response applications assisted by event-based non-intrusive load monitoring

With today's growth of prosumers and renewable energy resources, it is inevitable to incorporate the demand-side approaches for reliable and sustainable grid operation. In this context, demand response is a promising technique facilitating the consumers to play a substantial role in the energy market by altering their energy consumption patterns in times of peak demand or other critical contingencies. However, effective demand response deployment faces numerous challenges including trust deficit among the concerned stakeholders. This paper addresses the mentioned issue by proposing a non-invasive load-shed authentication model for demand response applications, assisted by an improved event-based non-intrusive load monitoring approach. For the said purposes, an improved event detection algorithm and machine learning model: support vector machine with a combination of genetic algorithm and GridSearchCV, is presented. This paper also presents a comprehensive real-world case study to validate the effectiveness of the proposed model in a real-life scenario. In the given context, all the simulations are carried out on low sampling real-world load measurements: Pecan Street-Dataport, where electric vehicle and air conditioning are employed as potential load elements for evaluation purposes. Based on the presented case study and analysis of the results, it is established that the presented improved event-based non-intrusive load monitoring approach yields promising performance in the context of multi-class classification. Moreover, it is also concluded that the proposed low sampling event-based non-intrusive load monitoring assisted non-invasive load-shed authentication model is a viable and promising solution for the effective implementation of demand response applications.

Simultaneous disaggregation of multiple appliances based on non-intrusive load monitoring

Non-intrusive load monitoring (NILM) can achieve the disaggregation of power consumption through electricity data at the power entrance without changing the existing circuit structure, which is helpful for saving energy, reducing emission, and improving the utilization of electric energy. This paper studies event-based NILM, which can disaggregate multiple appliances that are switched simultaneously. That is a mixed linear integer programming (MILP) model, where the 0–1 indicates whether the appliances were switched in the event process. And the 0–1 variable constraints of the power feature vectors during the event are constructed. The proposed MILP runs only when events occur, which can reduce the computational complexity. In event detection, non-dominated sorting genetic algorithm II is used to select the parameters of cumulative sum (CUSUM), such as threshold and the length of sliding windows, which realizes the tradeoff between precision and recall. The case studies demonstrate that the improved CUSUM has a precision of 92% in event detection; the proposed event-based NILM approach on the REDD database and laboratory data achieves an accuracy of over 90% on power draws; the appliance with long switching process, such as the inverter air conditioner, can also be monitored by the sequential combination.

Robust event-based non-intrusive appliance recognition using multi-scale wavelet packet tree and ensemble bagging tree

Providing the user with appliance-level consumption data is the core of each energy efficiency system. To that end, non-intrusive load monitoring is employed for extracting appliance specific consumption data at a low cost without the need of installing separate submeters for each electrical device. In this context, we propose in this paper a novel non-intrusive appliance recognition system based on (i) detecting events in the aggregated power signal using a novel and powerful scheme, (ii) applying multiscale wavelet packet tree to collect comprehensive energy consumption features, and (iii) adopting an ensemble bagging tree classifier along with comparing its performance with various machine learning schemes. Moreover, to validate the proposed model, an empirical investigation is conducted on two real and public energy consumption datasets, namely, the GREEND and REDD, in which consumption readings are collected at low-frequencies. In addition, a comprehensive review of recent non-intrusive load monitoring approaches has been conducted and presented, in which their characteristics, performances and limitations are described. The proposed non-intrusive load monitoring system shows a high appliance recognition performance in terms of the accuracy, F1 score and low time complexity when it has been applied to different households from the GREEND and REDD repositories, in which every house includes various domestic appliances. Obtained results have described, e.g., that average accuracies of 97.01% and 96.36% have been reached on the GREEND and REDD datasets, respectively, which outperformed almost existing solutions considered in this framework.

Robust energy disaggregation using appliance-specific temporal contextual information

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%.

Load Flexibility Forecast for DR Using Non-Intrusive Load Monitoring in the Residential Sector

Demand response services and energy communities are set to be vital in bringing citizens to the core of the energy transition. The success of load flexibility integration in the electricity market, provided by demand response services, will depend on a redesign or adaptation of the current regulatory framework, which so far only reaches large industrial electricity users. However, due to the high contribution of the residential sector to electricity consumption, there is huge potential when considering the aggregated load flexibility of this sector. Nevertheless, challenges remain in load flexibility estimation and attaining data integrity while respecting consumer privacy. This study presents a methodology to estimate such flexibility by integrating a non-intrusive load monitoring approach to load disaggregation algorithms in order to train a machine-learning model. We then apply a categorization of loads and develop flexibility criteria, targeting each load flexibility amplitude with a corresponding time. Two datasets, Residential Energy Disaggregation Dataset (REDD) and Refit, are used to simulate the flexibility for a specific household, applying it to a grid balancing event request. Two algorithms are used for load disaggregation, Combinatorial Optimization, and a Factorial Hidden Markov model, and the U.K. demand response Short Term Operating Reserve (STOR) program is used for market integration. Results show a maximum flexibility power of 200–245 W and 180–500 W for the REDD and Refit datasets, respectively. The accuracy metrics of the flexibility models are presented, and results are discussed considering market barriers.

Statistical and Electrical Features Evaluation for Electrical Appliances Energy Disaggregation

In this paper we evaluate several well-known and widely used machine learning algorithms for regression in the energy disaggregation task. Specifically, the Non-Intrusive Load Monitoring approach was considered and the K-Nearest-Neighbours, Support Vector Machines, Deep Neural Networks and Random Forest algorithms were evaluated across five datasets using seven different sets of statistical and electrical features. The experimental results demonstrated the importance of selecting both appropriate features and regression algorithms. Analysis on device level showed that linear devices can be disaggregated using statistical features, while for non-linear devices the use of electrical features significantly improves the disaggregation accuracy, as non-linear appliances have non-sinusoidal current draw and thus cannot be well parametrized only by their active power consumption. The best performance in terms of energy disaggregation accuracy was achieved by the Random Forest regression algorithm.

Automatic Time Series Segmentation as the Basis for Unsupervised, Non-Intrusive Load Monitoring of Machine Tools

Detailed energy monitoring and benchmarking at the individual component level is necessary to increase energy efficiency in complex production systems. Non-intrusive load monitoring (NILM) provides an economical solution for operational state detection and load disaggregation without the need for large-scale use of fine-grained energy meters. Existing supervised NILM approaches require detailed training data including control information about individual devices. Unsupervised approaches, on the other hand, often require high measurement resolution and are faced with the problem of detecting continuous states. This paper proposes a simple step-by-step, completely unsupervised NILM approach that distinguishes between almost constant and non-constant segments with flexible segment lengths. Taking into account various electrical parameters and their statistical moments, hierarchical density-based spatial clustering of applications with noise (HDBScan) is applied to constant segments. The analysis of non-constant segments is based on agglomerative hierarchical clustering and dynamic time warping. Based on real energy monitoring from a gear manufacturing system we show the applicability of our methodology and discuss how it can be combined with existing NILM techniques.

A Modified IP-Based NILM Approach Using Appliance Characteristics Extracted by 2-SAX

Deregulation on the delivery side of the power market has continuously been moving forward worldwide, which make bidirectional flow and interactions between customers and grids needs to be more refined and in-depth. Large-scale coverage of the advanced metering infrastructure (AMI) brings in skyrocketing of an immense amount of fine-grained, real-time consumption data and causes communication traffic congestion between meters and a cloud computing center. To tackle these two challenges, this paper proposes a modified IP-based non-intrusive load monitoring approach using appliance characteristics extracted by quadratic symbolic aggregate approximation (2-SAX). A 2-SAX algorithm is implemented to carry out dimensionality reduction on equipment load data and extracted the state's transition behavior characteristics and operation probability characteristics of each device. The extracted features can use to modify the disaggregation results of integer programming for overcoming the shortcomings of the previous IP approach. The developed method is tested with AMPds dataset. The results of experiments illustrate the 2-SAX consequences in 38.82%, 52.46%, and 13.41% reduction in MAE, MAPE, and RMSE on the heat pump and achieves similar performance on the other appliances, compared with normal SAX. Meanwhile, the proposed method MIP-AC2S delivers significant accuracy advantage and competitive performance over IP, ALIP, and MIP disaggregation method.

Electrical Energy Management Based on a Hybrid Artificial Neural Network-Particle Swarm Optimization-Integrated Two-Stage Non-Intrusive Load Monitoring Process in Smart Homes

Concerning electrical energy used in today’s modern society, electrical energy demands requested from downstream sectors in a smart grid are continuously increasing. One way to meet the electrical demands requested is to monitor and manage industrial, commercial, as well as residential electrical appliances efficiently in response to Demand Response (DR) programs for Demand-Side Management (DSM). Monitoring and managing electrical appliances that consume electrical energy in fields of interest can be realized through use of Energy Management Systems (EMS) with Non-Intrusive Load Monitoring (NILM). This paper presents an Internet of Things (IoT)-oriented Home EMS (HEMS). Also, a novel hybrid Artificial Neural Network-Particle Swarm Optimization (ANN-PSO)-integrated NILM approach is proposed and used to model and identify electrical appliances for DSM in the HEMS. ANN can be applied in NILM as a load identification task. Nevertheless, the performance of ANN used for load identification depends on three principal design factors: The network topology designed, the type of activation functions chosen, and the training algorithm adopted. As a result, PSO is conducted and used to incorporate meta-heuristics with ANN considering the three principal design factors relating to an ANN design. The HEMS with the novel hybrid ANN-PSO-integrated NILM proposed in this paper was deployed and evaluated in a realistic residential house environment. As the experimentation reported in this paper shows, the presented HEMS utilizing the proposed novel hybrid ANN-PSO-integrated NILM to model and identify monitored electrical appliances is feasible and workable, with an overall classification rate of 91.67% in load classification for DSM.

A NILM algorithm with enhanced disaggregation scheme under harmonic current vectors

In the smart home context, one of the features that most users would desire is the knowledge of the operation of each electrical appliance without the need to install expensive smart plugs or measuring equipment at each outlet-appliance. The Non-Intrusive Load Monitoring (NILM) idea is an alternative method where only one meter is utilized. However, the accuracy and simplicity of NILM approaches are still under improvement and evolvement. In this paper we propose the development of load signatures by a limited number of harmonic current vectors. The rationale behind this is that simultaneously operating appliances may have opposite or displaced harmonic current vectors, which means that using only current amplitudes as load signatures will not result in reliable identification. Three alternative formations are examined aiming to highlight the contribution of the harmonic current phase angles to the robustness of the load signatures. The first formation considers only the current amplitude, the second takes into account the phase angles in the form of vector projection on the x-axis, and the third is based on harmonic current vectors. The respective disaggregation schemes are also proposed considering either amplitude summations or vector aggregations, in order to examine the correct appliance combination that matches the actual operating one. Measurements have been performed in a typical residential installation considering both stand-alone operation of appliances and scenarios with appliance combinations. The results indicate that disaggregation performance is significantly improved when relying on harmonic current vectors in respect to the case with only current amplitudes, which makes the methodology potentially suitable for the new smart meters that are expected to be widely installed.

Enhancing neural non-intrusive load monitoring with generative adversarial networks

The application of Deep Learning methodologies to Non-Intrusive Load Monitoring (NILM) gave rise to a new family of Neural NILM approaches which increasingly outperform traditional NILM approaches. In this extended abstract describing our ongoing research, we analyze recent Neural NILM approaches and our findings imply that these approaches have difficulties in generating valid, reasonably-shaped appliance load profiles. We propose to enhance Neural NILM approaches with appliance load sequence generators trained with a Generative Adversarial Network to mitigate the described problem. The preliminary results of our experiments with Generative Adversarial Networks show the potential of the approach, albeit there is no strong evidence yet that this approach outperforms the examined end-to-end-trained Neural NILM approaches. In the progress of our investigations, we generalize energy-based NILM performance metrics and establish the complete classification confusion matrix based on the estimated energy in appliance load profiles. This enables the adaption of all known classification scores to their energy-based counterparts.

A generative model for non-Intrusive load monitoring in commercial buildings

In the recent years, there has been an increasing academic and industrial interest for analyzing the electrical consumption of commercial buildings. Whilst having similarities with the Non Intrusive Load Monitoring (NILM) tasks for residential buildings, the nature of the signals that are collected from large commercial buildings introduces additional difficulties to the NILM research causing existing NILM approaches to fail. On the other hand, the amount of publicly available datasets collected from commercial buildings is very limited, which makes the NILM research even more challenging for this type of large buildings. In this study, we aim at addressing these issues. We first present an extensive statistical analysis of both commercial and residential measurements from public and private datasets and show important differences. Secondly, we develop an algorithm for generating synthetic current data based on a modelization of the current flowing through an electrical device. We then demonstrate that our electrical device model fits well real measurements and that our simulations are realistic by using the quantitative metrics described in the previous section. Finally, to encourage research on commercial buildings we release a synthesized dataset called SHED that can be used to evaluate NILM algorithms.

A Supervised Event-Based Non-Intrusive Load Monitoring for Non-Linear Appliances

Smart meters generate a massive volume of energy consumption data which can be analyzed to recover some interesting and beneficial information. Non-intrusive load monitoring (NILM) is one important application fostered by the mass deployment of smart meters. This paper presents a supervised event-based NILM approach for non-linear appliance activities identification. Firstly, the additive properties (stating that, when a certain amount of specific appliances’ feature is added to their belonging network, an equal amount of change in the network’s feature can be observed) of three features (harmonic feature, voltage–current trajectory feature, and active–reactive–distortion (PQD) power curve features) were investigated through experiments. The results verify the good additive property for the harmonic features and Voltage–Current (U-I) trajectory features. In contrast, PQD power curve features have a poor additive property. Secondly, based on the verified additive property of harmonic current features and the representation of waveforms, a harmonic current features based approach is proposed for NILM, which includes two main processes: event detection and event classification. For event detection, a novel model is proposed based on the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Compared to other event detectors, the proposed event detector not only can detect both event timestamp and two adjacent steady states but also shows high detection accuracy over public dataset with F1-score up to 98.99%. Multi-layer perceptron (MLP) classifiers are then built for multi-class event classification using the harmonic current features and are trained using the data collected from the laboratory and the public dataset. The results show that the MLP classifiers have a good performance in classifying non-linear loads. Finally, the proposed harmonic current features based approach is tested in the laboratory through experiments, in which multiple on–off events of multiple appliances occur. The research indicates that clustering-based event detection algorithms are promising for future works in event-based NILM. Harmonic current features have perfect additive property, and MLP classifier using harmonic current features can accurately identify typical non-linear and resistive loads, which could be integrated with other approaches in the future.

Improving Event-Based Non-Intrusive Load Monitoring Using Graph Signal Processing

Large-scale smart energy metering deployment worldwide and integration of smart meters within the smart grid will enable two-way communication between the consumer and energy network, thus ensuring improved response to demand. Energy disaggregation or non-intrusive load monitoring (NILM), namely disaggregation of the total metered electricity consumption down to individual appliances using purely algorithmic tools, is gaining popularity as an added-value that makes the most of meter data. However, NILM remains a challenging problem since NILM is susceptible to sensor noise, unknown load noise, transient spikes, and fluctuations. In this paper, we tackle this problem using novel graph signal processing (GSP) concepts, applied at both, physical signal level via graph-based filtering and data level, via effective semi-supervised GSP-based feature matching. The proposed GSP-based method is generic and can be used to improve results of various event-based NILM approaches. We demonstrate significant improvement in performance using three state-of-the-art NILM methods, both supervised and unsupervised, and real-world active power consumption readings from the REDD and REFIT 1 data sets, sampled at 1 and 8 s, respectively. 1 The REFIT dataset used to generate the results can be accessed via DOI 10.15129/31da3ece-f902-4e95-a093-e0a9536983c4.

Non-intrusive Load Disaggregation Based on Kernel Density Estimation

Aiming at the problem of high cost and difficult implementation of high frequency non-intrusive load decomposition method, this paper proposes a new method based on kernel density estimation(KDE) for low frequency NILM (Non-intrusive load monitoring). The method establishes power reference model of electricity load in different working conditions and appliance’s possible combinations first, then probability distribution is calculated as appliances features by kernel density estimation. After that, target power data is divided by step changes, whose distributions will be compared with reference models, and the most similar reference model will be chosen as the decomposed consequence. The proposed approach was tested with data from the GREEND public data set, it showed better performance in terms of energy disaggregation accuracy compared with many traditional NILM approaches. Our results show good performance which can achieve more than 93% accuracy in simulation.

An error correction framework for sequences resulting from known state-transition models in Non-Intrusive Load Monitoring

Non-Intrusive Load Monitoring (NILM), the set of techniques used for disaggregating total electricity consumption in a building into its constituent electrical loads, has recently received renewed interest in the research community, partly due to the roll-out of smart metering technology worldwide. Event-based NILM approaches (i.e., those that are based on first segmenting the power time-series and associating each segment with the operation of electrical appliances) are a commonly implemented solution but are prone to the propagation of errors through the data processing pipeline. Thus, during energy estimation (the final step in the process), many corrections need to be made to account for errors incurred during segmentation, feature extraction and classification (the other steps typically present in event-based approaches). A robust framework for energy estimation should use the labels from classification to (1) model the different state transitions that can occur in an appliance; (2) account for any misclassifications by correcting event labels that violate the extracted model; and (3) accurately estimate the energy consumed by that appliance over a period of time. In this paper, we address the second problem by proposing an error-correcting algorithm which looks at sequences generated by Finite State Machines (FSMs) and corrects for errors in the sequence; errors are defined as state transitions that violate the said FSM. We evaluate our framework on simulated data and find that it improves energy estimation errors. We further test it on data from 43 appliances collected from 19 houses and find that the framework significantly improves errors in energy estimates when compared to the case with no correction in 19 appliances, leaves 17 appliances unchanged, and has a slightly negative impact on 6 appliances.

Smart sensor network for power quality monitoring in electrical installations

Abstract Smart meters are one of the basic components of the future smart grid, they allow remotely monitoring each point in the grid in order to know in real-time the performance of the system and to detect potential failures. In this paper, a smart sensor network is introduced and the most important features are presented in three different scenarios: a residential home, an industrial installation, and a public building. The proposed system demonstrates its capabilities of in situ real-time processing and big-data off-line network processing. The suggested smart meter is based on field programmable gate array (FPGA) technology that allows a reconfigurable architecture, which lets the user to select the proper processing modules according to their application. The developed smart sensor network calculates standard figures such as effective values, power factor, and total harmonic distortion; in addition, it detects power quality disturbances such as dips, swells, or interruptions. Moreover, the smart sensor network can continuously detect events to identify certain kind of appliances or industrial equipment such as: fans, lighting, microwave ovens, refrigerators, among others; it is a powerful tool to analyze an entire building in a non-intrusive load monitoring approach.

Context-Based Energy Disaggregation in Smart Homes

In this paper, we address the problem of energy conservation and optimization in residential environments by providing users with useful information to solicit a change in consumption behavior. Taking care to highly limit the costs of installation and management, our work proposes a Non-Intrusive Load Monitoring (NILM) approach, which consists of disaggregating the whole-house power consumption into the individual portions associated to each device. State of the art NILM algorithms need monitoring data sampled at high frequency, thus requiring high costs for data collection and management. In this paper, we propose an NILM approach that relaxes the requirements on monitoring data since it uses total active power measurements gathered at low frequency (about 1 Hz). The proposed approach is based on the use of Factorial Hidden Markov Models (FHMM) in conjunction with context information related to the user presence in the house and the hourly utilization of appliances. Through a set of tests, we investigated how the use of these additional context-awareness features could improve disaggregation results with respect to the basic FHMM algorithm. The tests have been performed by using Tracebase, an open dataset made of data gathered from real home environments.