This paper proposes and implements a smart architecture for Home Energy Management Systems (HEMS) that enables interoperability among devices from different manufacturers. This is achieved through the use of standardized elements and the design of an innovative middleware. The system comprises a control unit that communicates with smart outlets using the Wireless Smart Ubiquitous Network (WI-SUN) Home Area Network (HAN) specification, while smart metering is achieved using the WI-SUN Field Area Network (FAN) specification. To manage important data, a web platform and mobile app were created. Additionally, machine learning techniques are utilized to identify energy consumption of individual appliances when only the aggregate energy consumption of the house is available. The architecture presented here supports real-time control of energy use and generation through HEMS, and new devices can be added transparently. Finally, a comparison of the proposed system with similar systems in literature highlights its many advantages in terms of functionality.

A middleware software can be seem as an abstraction layer between hardware and user applications, that facilitates the development and deployment of services in various scenarios, such as those found in Home Energy Management Systems (HEMS). There are several middleware proposals for HEMS, with most of them taking the cloud computing approach. This approach is unconcerned about computing resources but raises a dependency on external connections. This paper presents a middleware for energy management systems, based on the concept of edge computing for smart homes. The paper presents a reference model for the proposed architecture, considering specific requirements for this type of application. The proposed architecture employs the concept of microservices for data access and system configuration. The proposed middleware is designed to work with embedded systems under computational constraints, such as processing capability and storage, to reduce costs and allow its application closer to the user. The middleware is open and customizable to meet the developer’s needs. The proposed solution was implemented and tested in a university laboratory, as well as at the Eldorado Research Institute to confirm the effectiveness of the middleware. The proposal stands out from others found in the literature as it can be implemented using low cost hardware. In addition to using microservices concepts, the proposed middleware is a valuable option for applications that need an edge computing approach. A performance analysis was carried out, using low cost hardware with limited resources. The results show that the proposal can handle a significant number of devices, offering low latency and low error rate, and consuming few processing resources and memory.

Instabilities in energy supply caused by equipment failures, particularly in power transformers, can significantly impact efficiency and lead to shutdowns, which can affect the population. To address this, researchers have developed fault diagnosis strategies for oil-immersed power transformers using dissolved gas analysis (DGA) to enhance reliability and environmental responsibility. However, the fault diagnosis of oil-immersed power transformers has not been exhaustively investigated. There are gaps related to real scenarios with imbalanced datasets, such as the reliability and robustness of fault diagnosis modules. Strategies with more robust models increase the overall performance of the entire system. To address this issue, we propose a novel approach based on Kolmogorov–Arnold Network (KAN) for the fault diagnosis of power transformers. Our work is the first to employ a dedicated KAN in an imbalanced data real-world scenario, named KAN_Diag, while also applying the synthetic minority based on probabilistic distribution (SyMProD) technique for balancing the data in the fault diagnosis. Our findings reveal that this pioneering employment of KAN_Diag achieved the minimal value of Hamming loss—0.0323—which minimized the classification error, guaranteeing enhanced reliability for the whole system. This ground-breaking implementation of KAN_Diag achieved the highest value of weighted average F₁-Score—96.8455%—ensuring the solidity of the approach in the real imbalanced data scenario. In addition, KAN_Diag gave the highest value for accuracy—96.7728%—demonstrating the robustness of the entire system. Some key outcomes revealed gains of 68.61 percentage points for KAN_Diag in the fault diagnosis. These advancements emphasize the efficiency and robustness of the proposed system.

The increasing domestic demand for electricity has required the scientific community to develop new technologies focused on energy expenditure minimization. In this context, Home Energy Management System (HEMS) is an element that can incorporate applications that aid in efficient energy management, such as load recognition. Load recognition allows HEMS to map all appliances, assisting in efficient management. In this regard, methods that solve the load recognition problem are vital for HEMS. The motivation of this work is to propose a novel system that enhances load recognition in HEMS applications. For this purpose, the proposed system contains an event detector stage based on the Wavelet transform and another stage for appliance identification consisting of two processing chains, one for high reliability and one for fast model training. The main novelties of our work are: (i) very-low training time via fast training processing chain, less than 1 second with k -Nearest Neighbor (k-NN); (ii) the highest values for accuracy, F ₁ , and the kappa index, 98.90%, 98.77%, and 0.9382, respectively, for the regularized Convolutional Neural Network (CNN-r) using the ‘Personalised Retrofit Decision Support Tools for UK Homes using Smart Home Technology’ (REFIT) dataset; (iii) this is the first paper employing CNN-r and Vision Transformer (ViT) for load recognition problem; and (iv) the approach is applicable in households without any restrictions, as long as training the models. Regarding the tested scenarios, the results reveal a very efficient strategy yielding accuracy values greater than 98% for the REFIT dataset and the Reference Energy Disaggregation Dataset (REDD). Finally, as presented in Table 1, the proposed approach provides the highest performance compared to other methods available in the literature.

This work investigates the efficiency of a fault detection system for refrigeration equipment based on Wavelet Transform. To perform the task, we used the well–known database REFIT (Personalised Retrofit Decision Support Tools for UK Homes using Smart Home Technology). We used data from six devices presented in three households of the database considering readings of active power as features. The results revealed an approach with high efficiency in detecting abnormal behavior of the tested appliances and presented low runtime despite large volumes of input data. Another positive point to highlight, in the proposed method, is the capacity to estimate the instants at which faults start and cease to occur. Considering the viewpoint of consumers, obtaining knowledge about the malfunctioning of household appliances enables timely repairs, resulting in financial savings and improved overall quality of life. Moreover, by setting parameters according to the characteristics of the appliances, the proposed method is suitable for different equipment, thereby establishing feasible fault detection for diverse groups of devices.

The dissolved gas analysis of insulating oil in power transformers can provide valuable information about fault diagnosis. Power transformer datasets are often imbalanced, worsening the performance of machine learning-based fault classifiers. A critical step is choosing the proper evaluation metric to select features, models, and oversampling techniques. However, no clear-cut, thorough guidance on that choice is available to date. In this work, we shed light on this subject by introducing new tailored evaluation metrics. Our results and discussions bring fresh insights into which learning setups are more effective for imbalanced datasets.