Sparse coding-based transfer learning for energy disaggregation

Deep attention and generative neural networks for nonintrusive load monitoring

Energy disaggregation algorithms disintegrate aggregate demand into appliance-level demands. Among various energy disaggregation approaches, non-intrusive load monitoring (NILM) algorithms requiring a single sensor have gained much attention in recent years. Various machine learning and optimization-based NILM approaches are available in the literature, but bulk training data and high computational time are their respective drawbacks. Considering these drawbacks, we devised an event matching energy disaggregation algorithm (EMEDA) for NILM of multistate household appliances using smart meter data. Having limited training data, K-means clustering was employed to estimate appliance power states. These power states were accumulated to generate an event database (EVD) containing all combinations of appliance operations in their various states. Prior to matching, the test samples of aggregate demand events were decreased by event-driven data compression for computational effectiveness. The compressed test events were matched in the sorted EVD to assess the contribution of each appliance in the aggregate demand. To counter the effects of transient spikes and/or dips that occurred during the state transition of appliances, a post-processing algorithm was also developed. The proposed approach was validated using the low-rate data of the Reference Energy Disaggregation Dataset (REDD). With better energy disaggregation performance, the proposed EMEDA exhibited reductions of 97.5 and 61.7% in computational time compared with the recent smart event-based optimization and optimization-based load disaggregation approaches, respectively.

Impulsive wavelet based probability sparse coding model for bearing fault diagnosis

Non-negative and local sparse coding based on l2-norm and Hessian regularization

Fisher discrimination dictionary sparse learning (FDDL) has led to interesting image recognition results where the Fisher discrimination criterion is subject to the coding coefficients. But Fisher discrimination criterion has the limitations of data distribution assumptions and does not consider the local manifold structure of the coding coefficients. In this paper, we will introduce a novel Fisher discrimination sparse learning based on graph embedding (GE-FDSL) scheme. First, we utilizes graph embedding framework to define intra-class compact matrix and inter-class separable matrix imposed on the coding coefficients of training samples to preserving the intra-class compactness and the inter-class separability for the training samples, which simultaneously consider the local manifold structure and label information of the coding coefficients. Then, a new Fisher discrimination criterion based on graph embedding is added to the object function of the sparse coding problem so that the coding coefficients have more discriminative power, where the dictionary atoms in the sparse coding model are associated with the class labels so that the reconstructed error is applied to classification. This method can learn a structured dictionary and sparse coefficients, and in the meantime, it will also keep the local manifold structure of the coding coefficients. So, they will be more discriminative. Experiments on many image databases show that the our algorithm has good classification and recognition performance.

Energy disaggregation technology is a key technology to realise real-time non-intrusive load monitoring (NILM). Current energy disaggregation methods use the same scale to extract features from the sequence, which makes part of the local features lost, resulting in low recognition accuracy of electrical appliances. Aiming at the low accuracy of non-intrusive energy disaggregation with low-frequency sampling, a sequential multi-scale residual network is proposed. Multi-scale residual network extracts multi-scale feature information through multi-scale convolution, and uses residual learning to deepen the network structure to further improve the performance of the algorithm. Different scales features are captured by multi-scale convolution to improve the recognition accuracy of electrical appliances with low using frequency. Sequence-to-sequence energy disaggregation mechanism can improve the disaggregation efficiency of the algorithm. The experimental comparison results show that the model can get better disaggregation effect, and can effectively identify the start-stop state of electrical appliances.

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 REFIT1 data sets, sampled at 1 and 8 s, respectively.

Energy disaggregation or Non-Intrusive Load Monitoring (NILM) addresses the issue of extracting device-level energy consumption information by monitoring the aggregated signal at one single measurement point without installing meters on each individual device. Energy disaggregation can be formulated as a source separation problem where the aggregated signal is expressed as linear combination of basis vectors in a matrix factorization framework. In this paper, an approach based on Sum-to-k constrained Non-negative Matrix Factorization (S2K-NMF) is proposed. By imposing the sum-to-k constraint and the non-negative constraint, S2K-NMF is able to effectively extract perceptually meaningful sources from complex mixtures. The strength of the proposed algorithm is demonstrated through two sets of experiments: Energy disaggregation in a residential smart home, and HVAC components energy monitoring in an industrial building testbed maintained at the Oak Ridge National Laboratory (ORNL). Extensive experimental results demonstrate the superior performance of S2K-NMF as compared to state-of-the-art decomposition-based disaggregation algorithms. The source code and our collected data (HVORUT) for studying NILM for HVAC units can be found at https://bitbucket.org/aicip/nonintrusive-load-monitoring.

Sparse coding with morphology segmentation and multi-label fusion for hyperspectral image super-resolution

Defect classification is an important part of automated surface defect detection systems. Generally, the types of surface defects of industrial products are complex and diverse, and the correct defect type classification can help to subsequently extract the characteristic of defects. Moreover, defect classification can help to count the number of defect modes automatically and provide data support for the precise maintenance of product production. Due to the large number of surface defects and the small difference between defect types, traditional classification methods are difficult to classify defect accurately. Therefore, in order to improve the accuracy of defect image classification, this paper proposes a defect image classification method based on transfer learning and sparse coding. Firstly, a deep CNN feature extraction algorithm for defect images is proposed in combination with transfer learning. Then, the deep CNN features of the defect image are dimension-reduced and sparsely optimized using the sparse coding techniques, and the sparse CNN features are obtained. Finally, the sparse CNN features are classified to realize the defect type determination using the linear SVM. The accuracy of the proposed method is verified by using a steel surface defect image benchmark database, and the effectiveness of the proposed method is proved.

Sequence to point learning based on bidirectional dilated residual network for non-intrusive load monitoring

A Novel Two-Dimensional Tensor Sparse Coding Algorithm for Image Representation

Multivariate exploration of non-intrusive load monitoring via spatiotemporal pattern network

Energy disaggregation (or Non-Intrusive Load Monitoring (NILM)) is the process of deducing individual load profiles from aggregate measurements using different machine learning and pattern recognition tools. Existing disaggregation algorithms can be categorized into either supervised approaches or unsupervised ones. Supervised approaches require external information represented in either sub-metered loads or hand-labeled observations while unsupervised algorithms utilize only unlabeled aggregate data. We observed that very few works attempt to utilize both labeled and unlabeled data. In this paper, we introduce a semi-supervised learning tool, namely self-training, to the energy disaggregation problem. Semi-Supervised Learning (SSL) tools leverage both external and internal structural information in order to enhance the learning process and/or reduce the required labeling effort. We also provide test results of the utilized SSL tool compared with a traditional classification component of an event-based NILM system. Results show that even a simple SSL tool is able to reduce the required labeling effort and provides a learning disaggregation system whose performance gradually increases as it observes more unlabeled aggregate measurements.

Non-intrusive load monitoring (NILM) is an important topic in smart-grid and\nsmart-home. Many energy disaggregation algorithms have been proposed to detect\nvarious individual appliances from one aggregated signal observation. However,\nfew works studied the energy disaggregation of plug-in electric vehicle (EV)\ncharging in the residential environment since EVs charging at home has emerged\nonly recently. Recent studies showed that EV charging has a large impact on\nsmart-grid especially in summer. Therefore, EV charging monitoring has become a\nmore important and urgent missing piece in energy disaggregation. In this\npaper, we present a novel method to disaggregate EV charging signals from\naggregated real power signals. The proposed method can effectively mitigate\ninterference coming from air-conditioner (AC), enabling accurate EV charging\ndetection and energy estimation under the presence of AC power signals.\nBesides, the proposed algorithm requires no training, demands a light\ncomputational load, delivers high estimation accuracy, and works well for data\nrecorded at the low sampling rate 1/60 Hz. When the algorithm is tested on\nreal-world data recorded from 11 houses over about a whole year (total 125\nmonths worth of data), the averaged error in estimating energy consumption of\nEV charging is 15.7 kwh/month (while the true averaged energy consumption of EV\ncharging is 208.5 kwh/month), and the averaged normalized mean square error in\ndisaggregating EV charging load signals is 0.19.\n