Non-intrusive load monitoring (NILM) provides fine-grained consumption information at the appliance level by analyzing the terminal voltage and total current measured. It shows prospective applications in demand side management, such as demand response, energy efficiency, and home energy management system. However, most cutting-edge NILM models have a critical assumption that switching events are triggered by known appliances in the training set, which may be unrealistic. In reality, new appliances are constantly being added, reducing the effectiveness of current methods for load monitoring. In this paper, we propose a conditional generative adversarial network (CGAN) to correctly classify all the known appliances while simultaneously detecting unknown ones using the V-I trajectory features. We integrate variational autoencoder and capsule networks in the generator network, the capsule features of the same known appliance class are forced to match a pre-defined Gaussian and a group of Gaussian priors (one for each appliance class) as the posterior distribution approximately. Furthermore, using an additional encoder network maps the generated V-I trajectory to its latent representation, minimizing the distance between the latent representations and the feature vectors of the generator aids in learning the data distribution for known appliance samples during the training process. In these ways, we can improve the ability of the model to detect unknown appliances. Experimental results on two public datasets demonstrate the effectiveness and superiority of our method.

Non-intrusive load monitoring (NILM) offers a practical solution for intelligent energy management in consumer electronics by estimating fine-grained energy consumption information for individual appliances from aggregated power signals measured at limited locations. With the rapid increase in appliance brands and types, constructing unique load signatures and establishing reliable appliance identification methods become progressively crucial for NILM. Deep neural networks have proven effective feature extraction capabilities and excellent identification performance in different fields. The primary approach in the current literature to extend this trend involves expanding the size of networks. However, this leads to rapidly increasing computational costs, with only marginal improvements in performance. In this paper, we propose and compare three different strategies for the fusion of the V-I trajectory and statistical features in an end-to-end trainable manner, evaluating feature fusion at various stages within the model (Feature-Level Fusion and Decision-Level Fusion), which automatically learn jointly from V-I features and a set of manually defined statistical features. As a result, the proposed approach improves cost efficiency while enhancing appliance identification performance. Experiments on diverse benchmark datasets demonstrate that the proposed feature fusion models cost-effectively achieve cutting-edge appliance identification performance.

Non-intrusive load monitoring (NILM), offers a privacy-preserving, real-time, and cost-effective approach to perceive power consumption patterns of appliances. The performance of NILM heavily relies on the quality of features extracted from load profiles. The shapelets method captures interpretable features that represent the unique characteristics of loads, proving highly relevant to NILM applications. In this work, an improved shapelets method for event-based NILM is proposed. First, a multi-length shapelets are creatively adopted for load identification, guaranteeing performance with lowfrequency or limited samples. Additionally, dynamic shapelets are obtained using a generative adversarial networks (GAN) instead of traditional search methods, ensuring efficient and distinctive shapelets generation across diverse appliances. Then, these generated shapelets are converted to distance features via shapelet transform and fed into classifier with squeeze-andexcitation network to enhance the accuracy of load identification. Experimental results on COOLL and BLUED datasets demonstrate that the proposed GAN-based multi-length dynamic shapelets method significantly optimizes the shapelet generation process and enhances the effectiveness of load identification in NILM.

Non-intrusive load monitoring (NILM) is a technique that disaggregates household total power consumption into the usage of individual appliances, which can promote the application of home energy management through detailed monitoring of electricity usage behavior. The rapid development of deep learning has driven significant advancements in the theory and algorithms of NILM. However, most deep learning models are fed with data from multiple households during the training process, and they suffer from performance decay due to the distribution discrepancy between the source domains and target domain. To address this issue, this paper proposes a novel data distillation-based multi-source domain adaptation mechanism that not only considers the different distribution distances between multiple source domains and the target domain but also investigates the varying similarities between source domain samples and target domain samples. Concretely, the proposed mechanism consists of four stages: a) pre-train the feature extractor and energy regressor using labeled data from the source domain; b) adversarially train the feature extractor that adaptive to the target domain by minimizing the empirical Wasserstein distance between the source and target domain; c) select the source domain samples that are closer to the target domain to fine-tune the energy regressor; d) aggregate power estimations from each source domain to the target domain based on each domain weight. Using data from five households in the public REFIT dataset, multiple comparative experiments validate the effectiveness of the proposed algorithm. Extensive experiments, including feature visualization and ablation studies, demonstrate that the proposed algorithm can significantly enhance the domain adaptation capabilities of the model.

Non-Intrusive Load Monitoring (NILM) provides a solution for smart energy management in consumer electronics, enabling the monitoring of internal appliance operating statuses by analyzing the aggregated load data. Recently, various machine learning algorithms have been employed for NILM, however, their effectiveness largely depends on the availability of substantial high-quality labeled data. In practice, the manual and automatic labeling based on event waveform extracted from master meter data inevitably introduce the noisy labels, which impacts the performance of load identification neural network model. To this end, we propose a label noise robust learning approach for non-intrusive load monitoring based on meta-learning. On the basis of traditional teacher-student framework, multiple meta networks are generated by adjusting student network parameters using pseudo labels, which are derived from the noisy labeled data according to neighbor feature similarity. These meta networks further generate predictions, which are compared with the teacher network predictions to constitute the meta loss. The NILM model trained with meta loss can extract valuable information from noisy labeled data and avoid overfitting to incorrect labels. Comparison tests on two public datasets and one private dataset demonstrate that our method provides better performance under different noise rates.