• Energy Research

The rising demand for energy requires high investments in network extensions and renewable sources, alongside replacing inefficient systems. Smart maintenance is important in minimizing unscheduled outages, reducing costs, improving network security, and increasing equipment's life expectancy. The vast amount of data collected by sensors and measurements in energy networks makes it hard for humans to detect failures continuously. Thanks to recent breakthroughs in AI, the energy sector has boosted the use of intelligent algorithms in this field. Despite the widespread popularity and great results of machine learning (ML) models in many applications, they are mostly nevertheless considered "black boxes" as understanding their functionality and transparency in real-world applications is challenging. Explainable Artificial Intelligence (XAI) tackles this by making AI systems' decision-making processes transparent and interpretable. This review paper will not only make the roadmap clear but also ensure an in-depth awareness of the challenges, opportunities, and developments associated with this path by presenting two comprehensive taxonomies. Various XAI methods are compared; as an example, our findings show that SHAP offers high trustworthiness but is less suited for real-time use, while LIME provides faster solutions with lower trustworthiness. To the best of the authors' knowledge, this is the first survey that provides an overview of XAI methods for energy systems maintenance (ESM). It addresses challenges like integrating XAI with IoT-powered digital twins, balancing explainability with cybersecurity, and ensuring scalability while proposing solutions to enhance reliability and efficiency.

The rising demand for energy requires high investments in network extensions and renewable sources, alongside replacing inefficient systems. Smart maintenance is important in minimizing unscheduled outages, reducing costs, improving network security, and increasing equipment's life expectancy. The vast amount of data collected by sensors and measurements in energy networks makes it hard for humans to detect failures continuously. Thanks to recent breakthroughs in AI, the energy sector has boosted the use of intelligent algorithms in this field. Despite the widespread popularity and great results of machine learning (ML) models in many applications, they are mostly nevertheless considered "black boxes" as understanding their functionality and transparency in real-world applications is challenging. Explainable Artificial Intelligence (XAI) tackles this by making AI systems' decision-making processes transparent and interpretable. This review paper will not only make the roadmap clear but also ensure an in-depth awareness of the challenges, opportunities, and developments associated with this path by presenting two comprehensive taxonomies. Various XAI methods are compared; as an example, our findings show that SHAP offers high trustworthiness but is less suited for real-time use, while LIME provides faster solutions with lower trustworthiness. To the best of the authors' knowledge, this is the first survey that provides an overview of XAI methods for energy systems maintenance (ESM). It addresses challenges like integrating XAI with IoT-powered digital twins, balancing explainability with cybersecurity, and ensuring scalability while proposing solutions to enhance reliability and efficiency.

Abstract Over the last two decades, there has been a growing realization that the actual energy performances of many buildings fail to meet the original intent of building design. Faults in systems and equipment, incorrectly configured control systems and inappropriate operating procedures increase the energy consumption about 20% and therefore compromise the building energy performance. To improve the energy performance of buildings and to prevent occupant discomfort, adverse condition and critical event prediction plays an important role. The Adverse Condition and Critical Event Prediction Toolbox (ACCEPT) is a generic framework to compare and contrast methods that enable prediction of an adverse event, with low false alarm and missed detection rates. In this paper, ACCEPT is used for fault detection and prediction in a real building at the University of Southern Denmark. To make fault detection and prediction possible, machine learning methods such as Kernel Density Estimation (KDE), and Principal Component Analysis (PCA) are used. A new PCA–based method is developed for artificial fault generation. While the proposed method finds applications in different areas, it has been used primarily for analysis purposes in this work. The results are evaluated, discussed and compared with results from Canonical Variate Analysis (CVA) with KDE. The results show that ACCEPT is more powerful than CVA with KDE which is known to be one of the best multivariate data-driven techniques in particular, under dynamically changing operational conditions.

Abstract Over the last two decades, there has been a growing realization that the actual energy performances of many buildings fail to meet the original intent of building design. Faults in systems and equipment, incorrectly configured control systems and inappropriate operating procedures increase the energy consumption about 20% and therefore compromise the building energy performance. To improve the energy performance of buildings and to prevent occupant discomfort, adverse condition and critical event prediction plays an important role. The Adverse Condition and Critical Event Prediction Toolbox (ACCEPT) is a generic framework to compare and contrast methods that enable prediction of an adverse event, with low false alarm and missed detection rates. In this paper, ACCEPT is used for fault detection and prediction in a real building at the University of Southern Denmark. To make fault detection and prediction possible, machine learning methods such as Kernel Density Estimation (KDE), and Principal Component Analysis (PCA) are used. A new PCA–based method is developed for artificial fault generation. While the proposed method finds applications in different areas, it has been used primarily for analysis purposes in this work. The results are evaluated, discussed and compared with results from Canonical Variate Analysis (CVA) with KDE. The results show that ACCEPT is more powerful than CVA with KDE which is known to be one of the best multivariate data-driven techniques in particular, under dynamically changing operational conditions.

District heating is one of the main strategies for providing heat supply for buildings in cold climate countries. However, large parts of the current prefabricated pipes of these networks are reaching the end of their technical service life. Replacing these pipes is a time-consuming procedure and requires huge investments. Predictive Maintenance (PdM) is a promising strategy to deal with these situations and to optimize and prioritize maintenance activities. This paper surveys different PdM approaches considering difficulties in implementing PdMs in district heating networks (DHNs) compared with other energy sectors. It demonstrates that the PdM methodology is unique for each DHN according to the distinctive characteristics, environmental factors, heating resource type, available data, equipment, and other factors. To the best of the authors’ knowledge, this paper presents the first comprehensive review focused on various aspects of PdM strategies developed for DHNs, including data analytics, prediction models, and integrated technologies to facilitate the implementation of these strategies in DHNs. A thorough discussion on state-of-the-art technologies and real-world challenges for implementing PdM is presented, and potential research avenues are provided.

District heating is one of the main strategies for providing heat supply for buildings in cold climate countries. However, large parts of the current prefabricated pipes of these networks are reaching the end of their technical service life. Replacing these pipes is a time-consuming procedure and requires huge investments. Predictive Maintenance (PdM) is a promising strategy to deal with these situations and to optimize and prioritize maintenance activities. This paper surveys different PdM approaches considering difficulties in implementing PdMs in district heating networks (DHNs) compared with other energy sectors. It demonstrates that the PdM methodology is unique for each DHN according to the distinctive characteristics, environmental factors, heating resource type, available data, equipment, and other factors. To the best of the authors’ knowledge, this paper presents the first comprehensive review focused on various aspects of PdM strategies developed for DHNs, including data analytics, prediction models, and integrated technologies to facilitate the implementation of these strategies in DHNs. A thorough discussion on state-of-the-art technologies and real-world challenges for implementing PdM is presented, and potential research avenues are provided.

Abstract Buildings account for ca. 40% of the total energy consumption and ca. 20% of the total CO2 emissions. More effective and advanced control integrated into Building Management Systems (BMS) represents an opportunity to improve energy efficiency. The ease of availability of sensors technology and instrumentation within today's intelligent buildings enable collecting high quality data which could be used directly in data-based analysis and control methods. The area of data-based systems analysis and control is concentrating on developing analysis and control methods that rely on data collected from meters and sensors, and information obtained by data processing. This differs from the traditional model-based approaches that are based on mathematical models of systems. We propose and describe a data-driven controllability measure for discrete-time linear systems. The concept is developed within a data-based system analysis and control framework. Therefore, only measured data is used to obtain the proposed controllability measure. The proposed controllability measure not only shows if the system is controllable or not, but also reveals the level of controllability, which is the information its previous counterparts failed to provide. We use two illustrative examples to demonstrate the method, which also include an intelligent building.

Abstract Buildings account for ca. 40% of the total energy consumption and ca. 20% of the total CO2 emissions. More effective and advanced control integrated into Building Management Systems (BMS) represents an opportunity to improve energy efficiency. The ease of availability of sensors technology and instrumentation within today's intelligent buildings enable collecting high quality data which could be used directly in data-based analysis and control methods. The area of data-based systems analysis and control is concentrating on developing analysis and control methods that rely on data collected from meters and sensors, and information obtained by data processing. This differs from the traditional model-based approaches that are based on mathematical models of systems. We propose and describe a data-driven controllability measure for discrete-time linear systems. The concept is developed within a data-based system analysis and control framework. Therefore, only measured data is used to obtain the proposed controllability measure. The proposed controllability measure not only shows if the system is controllable or not, but also reveals the level of controllability, which is the information its previous counterparts failed to provide. We use two illustrative examples to demonstrate the method, which also include an intelligent building.