Non-intrusive Load Monitoring Method Research Articles

Disaggregation of electricity and heating consumption in commercial buildings with building automation system data

Although understanding end uses in large commercial and institutional buildings is of great utility for achieving energy efficiency, most new and existing buildings still lack adequate submetering except for large end uses. End use disaggregation methods offer an alternative to fulfill weaknesses in a metering network, especially in large commercial and institutional buildings. Unlike non-intrusive load monitoring (NILM) methods that mostly rely on high frequency power consumption data and focus on residential buildings, this paper introduces an end-use disaggregation approach using building automation system (BAS) data. The BAS data provide information about the operational status of main energy-consuming systems and equipment such as fans, pumps, air handling unit (AHU) heating coils, and variable air volume (VAV) terminal reheat coils. The proposed method uses a suite of multiple linear regression models estimated by the genetic algorithm and a least-square solver to disaggregate heating and electricity use data using BAS data as predictors. The method was demonstrated with data from an academic office building in Ottawa, Canada. The accuracy was evaluated by comparing the disaggregation results to ground truth energy use data acquired from dedicated submeters. The findings suggest that the method can accurately disaggregate building-level hourly electricity and heating energy use into following end use categories, for electricity: (a) occupant-controlled loads (lighting and plug loads), (b) distribution loads (the electricity used by the AHUs); and for heating: (a) energy used by the AHU heating coils, (b) perimeter heating devices (reheat coils/and radiators).

Residential energy flexibility characterization using non-intrusive load monitoring

To accelerate progress in building sustainability as well as to aid balance supply and demand in the presence of renewable energy generation, a tailored characterization method for the energy flexibility (EF) of buildings is needed. In this paper, a novel two-stage non-intrusive EF characterization method is proposed. In the first stage, unlike the previous studies in which an individual meter is installed on appliances to extract their consumption pattern, a novel unsupervised event-matching non-intrusive load monitoring method is utilized which is time and cost-effective. Moreover, previous research characterize the EF considering as early and as late as possible appliances’ start-time. However, the usage behavior of consumers affects the start-time of appliances. To tackle this issue, in the second stage of the proposed method, the usage behavior of consumers is taken into account for the EF characterization. The proposed method is verified in an individual building level and aggregated level including 50 residential buildings. The obtained results show that the proposed usage behavior-oriented method, characterizes the available aggregated EF with higher accuracy, without adding complexity to the system. These results can be used by aggregators to harness the available EF of buildings to flatten demand consumption by incentivizing potential consumers.

A non-intrusive load monitoring algorithm based on multiple features and decision fusion

With the large-scale deployment of smart meters and wide application of various machine learning algorithms, non-intrusive load monitoring (NILM) has attracted the attention of academia and industry. However, machine learning algorithms often suffer from high variability in load identification performance due to different features. In view of the poor generalization ability and low accuracy of load identification using an individual feature or classifier model, this paper proposes a novel NILM method that accomplishes deep feature fusion and classifier model fusion by an improved Dempster–Shafer (D–S) evidence theory. Firstly, we extract the power features, current harmonic features, and voltage–current (V–I) trajectory features from the input signals. Then, K-nearest neighbor (KNN), random forest (RF), and convolutional neural networks (CNN) are employed to identify load appliances using three individual features. Finally, the probability estimates of each classifier are transmitted to the aggregator for aggregation to obtain the final identification results by the improved D–S evidence theory. The experimental results on the PLAID dataset show that the proposed method can significantly improve identification performances.

Comparison between artificial neural network and random forest for effective disaggregation of building cooling load

Detailed monitoring of loads can provide sufficient information about buildings and help in improving the operation of energy systems. Non-intrusive load monitoring (NILM) has become popular owing to its low cost and lack of disturbance in the occupied space. In this study, NILM methods based on artificial neural network (ANN) and random forest (RF) are proposed and compared to obtain the sub-loads of cooling loads. The cooling loads are disaggregated into four sub-loads: occupant load, equipment load, fresh air load, and building envelope load. Results showed both NILM methods could achieve load disaggregation accurately, but the RF-based NILM method performs better than that based on ANN. Moreover, the RF-based NILM method has mean relative errors (MREs) between 2.5% and 13.0%, while the ANN-based NILM method has MREs between 3.1% and 12.9%. Among the four sub-loads, the equipment load could be disaggregated with the highest accuracy. The detailed sub-loads obtained using the proposed NILM methods can guide building renovation and optimization design of building energy systems.

Non-intrusive load monitoring method with inception structured CNN

Non-intrusive load monitoring method with inception structured CNN

Non-Intrusive Load Monitoring for Residential Appliances with Ultra-Sparse Sample and Real-Time Computation.

To achieve the goal of carbon neutrality, the demand for energy saving by the residential sector has witnessed a soaring increase. As a promising paradigm to monitor and manage residential loads, the existing studies on non-intrusive load monitoring (NILM) either lack the scalability of real-world cases or pay unaffordable attention to identification accuracy. This paper proposes a high accuracy, ultra-sparse sample, and real-time computation based NILM method for residential appliances. The method includes three steps: event detection, feature extraction and load identification. A wavelet decomposition based standard deviation multiple (WDSDM) is first proposed to empower event detection of appliances with complex starting processes. The results indicate a false detection rate of only one out of sixteen samples and a time consumption of only 0.77 s. In addition, an essential feature for NILM is introduced, namely the overshoot multiple (which facilitates an average identification improvement from 82.1% to 100% for similar appliances). Moreover, the combination of modified weighted K-nearest neighbors (KNN) and overshoot multiples achieves 100% appliance identification accuracy under a sampling frequency of 6.25 kHz with only one training sample. The proposed method sheds light on highly efficient, user friendly, scalable, and real-world implementable energy management systems in the expectable future.

A Non-intrusive loading monitoring method based on Attention mechanism and denoising-Autoencoder

This paper built denoising-Autoencoder (dAE), Attention-dAE and transformerbased regression models based on Non-Intrusive Load Monitoring (NILM) and Deep Learning (DL) methods, and uses them in UK-DALE[1] proving on the dataset, the verification results show that the recognition accuracy of adding Attention mechanism and dAE is significantly improved, and the model parameters are greatly reduced.

A Non-intrusive Load Identification Algorithm Combined with Event Detection

Background: The traditional state-based non-intrusive load monitoring method mainly deploys the aggregate load as the characteristic to identify the states of every electrical appliance. Each identification is relatively independent, and there is no correlation between the identification results. Objective: This paper combines the event detection results with the state-based non-intrusive load identification algorithm to improve accuracy. Methods: Firstly, the load recognition model based on an artificial neural network is constructed, and the state-based recognition results are obtained. An event recognition and detection model is then built to identify electrical state transitions, that is, the current moment based on the event recognition results obtained from the previous moment. Finally, a reasonable decision method is constructed to determine the identification result of the electrical states. Result: Experimental results on the public data set REDD show that in the Long Short-Term Memory (LSTM) fusion model, the macro-F1 is increased by an average of 6%, and the macro-F1 of the Artificial Neural Network (ANN) fusion model is increased by an average of 5.3% compared with LSTM and ANN. Conclusion: The proposed model can effectively improve the accuracy of identification compared with the state-based load identification method.

Cooling load disaggregation using a NILM method based on random forest for smart buildings

Accurate load monitoring can provide detailed information for users to improve the energy efficiency of buildings. Non-intrusive load monitoring (NILM) has become popular in smart buildings because of its low cost and reasonable privacy. In this paper, a non-intrusive monitoring method for the cooling load of smart buildings is proposed based on random forest. The total building cooling load is disaggregated into four subloads, and two approaches are used to realize the NILM based on the direct or indirect cooling load using a Fourier transform. The proposed method is implemented in an office building, and results show the method can realize cooling load disaggregation accurately. The root-mean-square errors and mean relative errors of the four subloads between the NILM loads and reference loads are less than 51.9 kW and 19.1%. Among the four subloads, the equipment load can be disaggregated with the highest accuracy. Approach I is recommended because of its higher accuracy. The NILM method is optimized in terms of the estimator number, maximum depth, feature number, minimum samples for a split, minimum sample leaf, and size of training samples. The performance of the optimized NILM models is improved with RMSEs and MREs less than 48.3 kW and 6.4%.

Toward Load Identification Based on the Hilbert Transform and Sequence to Sequence Long Short-Term Memory

Load identification is a core concept in non-intrusive load monitoring (NILM). Through NILM systems, users can check their home appliance usage habits and then adjust their behavior to save electricity. In this way, a NILM system offers an effective method to detect the event status of household appliances as well as individual loads' energy consumption. However, prior NILM methods have encountered a challenge in improving recognition accuracy for both linear load and non-linear load types. These methods used a representative feature, namely transient load signals. However, the transient signals on these loads differ in terms of transient time and transient shape, which is the main cause of reduced accuracy performance in load identification. To this end, this paper presents a novel method, HT-LSTM (Hilbert Transform Long Short-Term Memory), which enhances recognition of the various load types that contain the difference in the transient time and the transient shape of any load signal. The proposed method consists of two main parts: (i) generating a novel transient feature based on a Hilbert transform (HT), called APF (Amplitude-Phase-Frequency). APF features are sequential data, which is used for the classification model; and (ii) applying Sequence-to-Sequence Long Short-Term Memory (Seq2Seq LSTM) to identify appliances by using APF features as the input data. In this work, we evaluate the HT-LSTM method using two high-frequency public datasets, Building-Level fUlly-labeled dataset for Electricity Disaggregation (BLUED) and Plug Load Appliance Identification Dataset (PLAID). Also, we evaluate our method using a private dataset collected in the lab. Based on the experimental results obtained and comparison classification performance pointed, the proposed method outperforms previous methods of F-score measurement on both public datasets in load identification as well as the private dataset.

Domestic smart metering infrastructure and a method for home appliances identification using low‐rate power consumption data

The deployment of domestic smart metering infrastructure in Great Britain provides the opportunity for identification of home appliances utilising non-intrusive load monitoring methods. Identifying the energy consumption of certain home appliances generates useful insights for the energy suppliers and for other bodies with a vested interest in energy consumption. Consequently, the domestic smart metering system, which is an integral part of the smart cities' infrastructure, can also be used for home appliance identification purposes taking into account the limitations of the system. In this article, a step-by-step description on accessing data directly from the domestic Smart Meter via an external Consumer Access Device is described, as well as an easy-to-implement method for identifying commonly used home appliances through their power consumption signals sampled at a rate similar to the rate available by the domestic smart metering system. The experimental results indicate that the combination of time domain with frequency domain features extracted either from the 1D/2D Discrete Fourier Transform or the Discrete Cosine Transform provides improved recognition performance compared to the case where the time domain or the frequency domain features are used separately.

Event Matching Classification Method for Non-Intrusive Load Monitoring

Nowadays, energy management aims to propose different strategies to utilize available energy resources, resulting in sustainability of energy systems and development of smart sustainable cities. As an effective approach toward energy management, non-intrusive load monitoring (NILM), aims to infer the power profiles of appliances from the aggregated power signal via purely analytical methods. Existing NILM methods are susceptible to various issues such as the noise and transient spikes of the power signal, overshoots at the mode transition times, close consumption values by different appliances, and unavailability of a large training dataset. This paper proposes a novel event-based NILM classification algorithm mitigating these issues. The proposed algorithm (i) filters power signals and accurately detects all events; (ii) extracts specific features of appliances, such as operation modes and their respective power intervals, from their power signals in the training dataset; and (iii) labels with high accuracy each detected event of the aggregated signal with an appliance mode transition. The algorithm is validated using REDD with the results showing its effectiveness to accurately disaggregate low-frequency measured data by existing smart meters.

PUMPNET: a deep learning approach to pump operation detection

Non-urgent high energy-consuming residential appliances, such as pool pumps, may significantly affect the peak to average ratio (PAR) of energy demand in smart grids. Effective load monitoring is an important step to provide efficient demand response (DR) to PAR. In this paper, we focus on pool pump analytics and present a deep learning framework, PUMPNET, to identify the pool pump operation patterns from power consumption data. Different from conventional time-series based Non-intrusive Load Monitoring (NILM) methods, our approach transfers the time-series data into image-like (date-time matrix) data. Then a U-shaped fully convolutional neural network is developed to detect and segment the image-like data in pixel level for operation detection. Our approach identify whether pool pumps operate given thirty-minute interval aggregated active power consumption data in kilowatt-hours only. Furthermore, the PUMPNET algorithm could identify pool pump operation status with high accuracy in the low-frequency sampling scenario for thousands of household, compared to traditional NILM algorithms which process high sampling rate data and can only apply to limited number of households. Experiments on real-world data validate the promising results of the proposed PUMPNET model.

Realistic scenario modelling for building power supply and distribution system based on non-intrusive load monitoring

In power supply and distribution system in buildings, the conventional designs of loads are fictitious, and it is difficult to find the vulnerabilities timely. In order to solve this problem, a realistic scenario modelling method is proposed. Aiming at the input of the power consumption data in realistic scenario model, a non-intrusive load monitoring method is used, combined with sliding window switching event detection method of electrical appliance, a sequence to short-sequence deep learning model is also established whose input vectors are composed of switching time and total power data. The input vectors in the deep learning model can be decomposed to the individual electrical appliance. Compared with CO and FHMM algorithm, the decomposition results of this model are more excellent in precision, recall, F1 score and accuracy. And it is more practical and accurate to replace the estimated data with the electricity consumption data obtained by NILM in the realistic scenario modelling.

An Improved Load Feature Extraction Technique for Smart Homes Using Fuzzy-Based NILM

With the advancement of smart residential buildings, houseowners are increasingly interested to monitor the connected electrical loads in view of energy conservation and load management. Smart grid technology in the distribution system includes cutting-edge load monitoring applications, permitting efficient use of electrical energy. Nonintrusive load monitoring (NILM) technique is such a tool to recognize the consumption of energy for individual electrical loads in a residential system at a solitary point of measurement. In this article, an improved NILM method is proposed using a shunt passive filter installed at the source end of residential building with the help of current harmonic signature analysis. A fuzzy rule-based intelligent identification method is employed in order to detect and monitor different electrical loads considering various load features at the utility side. The major contribution of the proposed technique is the improvement in identification accuracy of household loads with multiple appliances. The connected filter additionally restricts the injected harmonics generated by the loads at the source end. The usage of the fuzzy technique also reduces the processor memory requirement than conventional machine learning techniques. Simulations backed by experimental results validate the viability of the proposed methodology.

Non-intrusive load monitoring based on harmonic characteristics

Abstract Non-intrusive Load Monitoring (NILM) technology is based only on the voltage at the total entrance of the power supply to the home and so on. The measurement and analysis of the current can obtain the electrical information of the electrical equipment in the scene. By detecting and classifying the load events, a NILM method based on the load events is proposed in the paper. The working state and power consumption of each electrical appliance in the scene can be inferred, and the non-intrusive load monitoring is realized. The method of edge detection was used to locate the load events generated by electrical equipment from the aggregated data. In order to improve the accuracy of monitoring, the voltage is increased on the basis of the power characteristics, and the support vector machine (SVM) was used as the classification model. Finally, the load events of various electrical appliances were collected and labeled in the field home data set. The accuracy of load event identification based on the classifier trained with different features is compared, and the effectiveness of the proposed method is verified.

A Framework to Generate and Label Datasets for Non-Intrusive Load Monitoring

In order to reduce the electricity consumption in our homes, a first step is to make the user aware of it. Raising such awareness, however, demands to pinpoint users of specific appliances that unnecessarily consume electricity. A retrofittable and scalable way to provide appliance-specific consumption is provided by Non-Intrusive Load Monitoring methods. These methods use a single electricity meter to record the aggregated consumption of all appliances and disaggregate it into the consumption of each individual appliance using advanced algorithms usually utilizing machine-learning approaches. Since these approaches are often supervised, labelled ground-truth data need to be collected in advance. Labeling on-phases of devices is already a tedious process, but, if further information about internal device states is required (e.g., intensity of an HVAC), manual post-processing quickly becomes infeasible. We propose a novel data collection and labeling framework for Non-Intrusive Load Monitoring. The framework is comprised of the hardware and software required to record and (semi-automatically) label the data. The hardware setup includes a smart-meter device to record aggregated consumption data and multiple socket meters to record appliance level data. Labeling is performed in a semi-automatic post-processing step guided by a graphical user interface, which reduced the labeling effort by 72% compared to a manual approach. We evaluated our framework and present the FIRED dataset. The dataset features uninterrupted, time synced aggregated, and individual device voltage and current waveforms with distinct state transition labels for a total of 101 days.

Non-Intrusive Monitoring Algorithm for Resident Loads with Similar Electrical Characteristic

Non-intrusive load monitoring is a vital part of an overall load management scheme. One major disadvantage of existing non-intrusive load monitoring methods is the difficulty to accurately identify loads with similar electrical characteristics. To overcome the various switching probability of loads with similar characteristics in a specific time period, a new non-intrusive load monitoring method is proposed in this paper which will modify monitoring results based on load switching probability distribution curve. Firstly, according to the addition theorem of load working currents, the complex current is decomposed into the independently working current of each load. Secondly, based on the load working current, the initial identification of load is achieved with current frequency domain components, and then the load switching times in each hour is counted due to the initial identified results. Thirdly, a back propagation (BP) neural network is trained by the counted results, the switching probability distribution curve of an identified load is fitted with the BP neural network. Finally, the load operation pattern is profiled according to the switching probability distribution curve, the load operation pattern is used to modify identification result. The effectiveness of the method is verified by the measured data. This approach combines the operation pattern of load to modify the identification results, which improves the ability to identify loads with similar electrical characteristics.

Non-Intrusive Load Monitoring via Deep Learning Based User Model and Appliance Group Model

Non-Intrusive Load Monitoring (NILM) increases awareness on user energy usage patterns. In this paper, an efficient and highly accurate NILM method is proposed featuring condensed representation, super-state and fusion of two deep learning based models. Condensed representation helps the two models perform more efficiently and preserve longer-term information, while super-state helps the model to learn correlations between appliances. The first model is a deep user model that learns user appliances usage patterns to predict the next appliance usage behavior based on past behaviors by capturing the dynamics of user behaviors history and appliances usage habits. The second model is a deep appliance group model that learns the characteristics of appliances with temporal and electrical information. These two models are then fused to perform NILM. The case study based on REFIT datasets demonstrates that the proposed NILM method outperforms two state-of-the-art benchmark methods.

On-line Detection of Malicious Activities Based on Edge Computing in Micro-grid System

With the continuous deployment of smart grids, various new smart technologies applied to the power grids have emerged, and the security boundaries of the power systems have gradually blurred, so that the power security protection measures urgently need to be updated. Aiming at the smart micro-grid system based on edge computing, this paper introduces a non-intrusive load monitoring (NILM) method, combined with the advantages of edge computing, and designs an online detection mechanism for malicious activities of terminal devices. This method is dedicated to achieving device-level security assurance.