Automating the task of fault detection and diagnosis is essential for reducing the operational and maintenance costs of power electronic converters. Hence, this paper introduces an online optimization methodology for digital twins to diagnose multiple faults in grid connected inverters. In addition to enhancing accuracy, we channelize our efforts towards reducing computational complexity in parameter configuration under a limited set of training data. To address cybersecurity vulnerabilities during the training phase of our digital twin, we performdata sanitizationwith the aid of a quantitative association rule mining technique. For classification performance assessment, various fault cases in a virtual synchronous generator are considered to demonstrate the efficacy of our approach. Our research outcomes reveal increased accuracy and fidelity levels achieved by our digital-twin design, paving the way for reliable and trustworthy decision-making under anomalous conditions.

As cybercrime is becoming increasingly sophisticated, effective cybersecurity is crucial to safeguard digital assets and protect critical infrastructures from emerging threats. Several security applications exploit recent advances in (Big) data analysis and Artificial Intelligence (AI) to prevent and respond to malicious activities. Towards this direction, supervised and unsupervised Machine Learning (ML) methods are used to detect anomalies or reveal patterns that may indicate potential threats. However, the successful implementation of these technologies requires security practitioners to undergo specialized training to fully understand and use AI-driven tools and data analytics. On the other hand, AI models themselves are vulnerable to a variety of cyber threats, which can compromise their training data and learning processes. To ensure the safe operation of these systems, especially when deployed in adversarial environments, it is crucial to create novel AI adversarial algorithms and models that are resilient against diverse security threats. This work presents a conceptual framework based on Large Language Models (LLMs) supported by a Multi-Agent layer for training of security practitioners in various advanced technologies and enhance ML models ability to detect and respond to emerging cyber threats effectively.

Abstract Learning methods have been increasingly used in power engineering to perform various tasks. In this paper, a fault selection procedure in double-circuit transmission lines employing different learning methods is accordingly proposed. In the proposed procedure, the discrete Fourier transform (DFT) is used to pre-process raw data from the transmission line before it is fed into the learning algorithm, which will detect and classify any fault based on a training period. The performance of different machine learning algorithms is then numerically compared through simulations. The comparison indicates that an artificial neural network (ANN) achieves remarkable accuracy of 98.47%. As a drawback, the ANN method cannot provide explainable results and is also not robust against noisy measurements. Subsequently, it is demonstrated that explainable results can be obtained with high accuracy by using rule-based learners such as the recently developed quantitative association rule mining algorithm (QARMA). The QARMA algorithm outperforms other explainable schemes, while attaining an accuracy of 98%. Besides, it was shown that QARMA leads to a very high accuracy of 97% for highly noisy data. The proposed method was also validated using data from an actual transmission line fault. In summary, the proposed two-step procedure using the DFT combined with either deep learning or rule-based algorithms can accurately and successfully perform fault selection tasks but indicating remarkable advantages of the QARMA due to its explainability and robustness against noise. Those aspects are extremely important if machine learning and other data-driven methods are to be employed in critical engineering applications.

Predictive Analytics and Machine Learning (ML) are at the heart of some of the most popular Industry 4.0 applications such as condition-based monitoring, predictive maintenance, and quality management. To support the implementation of such use cases, various ML models have been proposed and validated in the research literature. This paper introduces a novel set of machine learning algorithms for Industry4.0 use cases, namely the QARMA algorithms, which are capable of mining of quantitative rules. QARMA models present several advantages when compared to conventional ML and Deep Learning mechanisms, including computational performance, predictive accuracy and "explainability". In the scope of this paper, we discuss these advantages based on practical experiences from the field deployment and validation of QARMA models in two different production lines. The deployment has been supported by a state-of-the-art Industrial Internet of Things platform, which is also presented in the paper.

In this paper we are investigating the capacity of Mobile Ad Hoc Networks under a concurrent transmission constraint imposed by common MAC layer protocols (such as 802.11). This problem is known to be an NP-hard combinatorial problem and a greedy algorithm has been proposed for solving it in polynomial time. We propose an efficient algorithm for solving the exact problem and compare the exact and approximate solutions for a variety of network sizes and node distributions (resulting from commonly used mobility models). Our findings confirm that the greedy solution provides a good approximation of the exact solution even for small network sizes. Based on the exact solution, we also derive an empirical formula that relates the capacity optimizing transmission range to the number of wireless nodes uniformly distributed in a square unit area.

We present a novel optimization-based method for the combination of cluster ensembles for the class of problems with intracluster criteria, such as Minimum-Sum-of-Squares-Clustering (MSSC). We propose a simple and efficient algorithm-called EXAMCE-for this class of problems that is inspired from a Set-Partitioning formulation of the original clustering problem. We prove some theoretical properties of the solutions produced by our algorithm, and in particular that, under general assumptions, though the algorithm recombines solution fragments so as to find the solution of a Set-Covering relaxation of the original formulation, it is guaranteed to find better solutions than the ones in the ensemble. For the MSSC problem in particular, a prototype implementation of our algorithm found a new better solution than the previously best known for 21 of the test instances of the 40-instance TSPLIB benchmark data sets used in [1], [2], and [3], and found a worse-quality solution than the best known only five times. For other published benchmark data sets where the optimal MSSC solution is known, we match them. The algorithm is particularly effective when the number of clusters is large, in which case it is able to escape the local minima found by K-means type algorithms by recombining the solutions in a Set-Covering context. We also establish the stability of the algorithm with extensive computational experiments, by showing that multiple runs of EXAMCE for the same clustering problem instance produce high-quality solutions whose Adjusted Rand Index is consistently above 0.95. Finally, in experiments utilizing external criteria to compute the validity of clustering, EXAMCE is capable of producing high-quality results that are comparable in quality to those of the best known clustering algorithms.

We report on lessons learnt from operating a location-aware mobile social networking application, and critical functionalities that are needed in order to provide a desirable user experience. As a result of user feedback, we enhanced our social networking system, G2G, with functionalities such as login with Facebook.com credentials without the need to sign up to our system first, and a much improved localization system that works across different mobile operators. These functionalities, combined with the ability of the system to upload/search/download location-aware multimedia notes, resulted in a significantly enhanced user experience as evidenced by higher user active participation in the system.

In this paper, we propose a novel hybrid system for handling reputation in super-peer-based networks by combining the personal history of each user's interactions with other users, the opinions of peer-friends together with global ratings of peers as they emerge from all of their interactions with other users of the network. We introduce the notion of peer friends in a P2P network and use it to prevent malicious collectives from reducing the reputation of a peer in the network. We also present a secure distributed framework that ensures that trust reports remain encrypted and are never opened during the submission or aggregation process. Computational results from our distributed prototype simulation show that our solution compares favorably with all other proposed methods for handling reputation when subject to various malicious strategies.

We present G2G, a proximity-aware social networking platform for Mobile Information Device (MID) users and its applications in personalized recommender systems and gaming. User proximity to a friend or a place is determined via GSM cell information. GSM cell information is greatly enhanced through the concept of user-defined "hotspots" which is the means that allows the system to expand the notion of proximity to include not only users covered by the same GSM cell but also users and/or hotspots covered by geographically neighboring cells making it one of the main innovations of our system. Overall, the system allows users to see nearby friends and chat with them, or leave notes and/or recommendations valid only in their current location. The system's flexibility allows for users to set-up their own treasurehunt games and other applications.

In this paper, we describe SuperTrust, a novel and efficient framework designed to handle trust relationships in Super-peer networks. What distinguishes SuperTrust from other works is that trust reports remain encrypted and are never opened during the submission or aggregation processes, thus guaranteeing privacy, anonymity, fairness, persistence and eligibility of transactions.

The authors presented a hyper-graph partitioning based approach to the problem of load-balancing in large-scale P2P-based digital libraries. In such libraries, in addition to publishing digital content, peer-clients may subscribe their intent to be notified whenever digital content matching appropriate meta-data keywords is published in the library (a.k.a. continuous queries or selective dissemination of information), as well as perform more traditional point queries. The digital content, as well as the continuous queries, are partitioned among a number of blocks, and each super-peer server in the network is responsible for handling only one of these partitioned blocks. Preliminary results showing that the hyper-graph partitioning approach scales well with the number of published documents or continuous queries, and with the size of the network were presented. The results show that partitioning-based load balancing of the super-peers leads to significant improvements of the average response time for point or continuous queries, and much reduced storage requirements of the library metadata.