Publication Date: 2026
Engineering Applications of Artificial Intelligence (09521976)168
As the global transition toward intelligent, resilient energy infrastructures accelerates, integrating artificial intelligence (AI) into power system protection has become a critical enabler of operational efficiency and reliability. A major challenge is the accurate identification and classification of high impedance faults (HIFs) in energy distribution grids (EDGs), where low current levels often cause conventional protection devices to malfunction. Traditional schemes require extensive training data, which is often difficult or impossible to obtain. This study presents a novel protection scheme that combines a pre-trained compact convolutional neural network (SqueezeNet) with Wigner-Ville distribution (WVD) and S-transform-based feature fusion to detect and classify HIFs using minimal data. Leveraging transfer learning, the approach reduces model retraining needs and accelerates deployment. Simulation results on modified IEEE 13-bus and 34-bus EDGs show F 1-scores exceeding 97 %, successful cross-network knowledge transfer without retraining, and rapid detection within 20 msec using only 250 training samples, highlighting its suitability for lightweight, scalable, and real-time smart-grid protection. © 2026 Elsevier Ltd.
Publication Date: 2022
Physica A: Statistical Mechanics and its Applications (03784371)604
Community detection is one of the most essential issues in social networks analysis field. Among the available categories of algorithms, the label propagation algorithm-based (LPA-based) methods, due to their proper time complexity, are of high concern. As all the social networks explicitly or implicitly include signed relationships, the attempt here is to suggest an LPA-based approach for community detection in the directed signed social networks. The direction of edges is not addressed in available LPA-based community detection methods for signed social networks. In this respect, 1) a weighting method is suggested in order to utilize the direction information that converts the network into an undirected weighted signed social network, 2) this weight is combined with a second weight obtained from the sign information of the edges, and 3) the LPA is extended, where the combined weights are applied in label propagation. Moreover, the directed signed modularity and the directed signed flow-based capacity measures are proposed. The findings of the run experiments indicate that the proposed method as to the directed signed modularity, directed signed flow-based capacity, and frustration measures on real-world and synthetic data sets, outperforms its counterparts. © 2022 Elsevier B.V.
Publication Date: 2024
IET Generation, Transmission and Distribution (17518687)18(4)pp. 767-778
Power transformers play a critical role in the performance of power systems. This equipment is costly due to significant copper and iron prices and manufacturing costs. Therefore, maintenance and protection of such equipment is essential. Despite its robust performance, maloperation of differential protection (DP) in transformers may cause operational challenges to power system operators. The differential relay may operate incorrectly after the transformer energization leads to an inrush current (IC) and the relay identifies the event as an internal fault, and consequently issues the trip command. The other case of maloperation includes, but not limited to, a moment when the current transformer saturates due to an external fault. In this paper, a novel approach for DP is proposed, that is based on signal processing methods. In this paper, variational mode decomposition (VMD) and the deep neural network are implemented by using the convolutional neural network (CNN) and bi-directional long short-term memory (BiLSTM) models. The VMD decomposes differential current signal (DCS) to intrinsic mode functions with corresponding narrow-band property frequency spectrums, which provides more detailed information about signal characteristics in different frequency bands. At the next stage, an effective feature for the BiLSTM is extracted by the CNN with the convolutional layers to classify events and proper discrimination. Extensive simulations on a 500 MVA transformer in MATLAB demonstrate the effectiveness of the proposed protection approach to differentiate ICs from internal and external faults with 99.8% accuracy in less than 1/8th of a power cycle. © 2024 The Authors. IET Generation, Transmission & Distribution published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.
Publication Date: 2022
Electric Power Systems Research (03787796)212
Low level current and similarity of High Impedance Faults (HIF) in respect of characteristics to other transient events have posed a critical challenge to the protection of distribution systems. In addition, the dependency of previous methods on large amounts of training data increases the simulation error rate, and preparing this amount of data is time-consuming. In this paper, a novel scheme based on conditional generative adversarial network (CGAN) and convolutional neural network (CNN) classifier techniques is proposed, that reduces this dependency and leads to acceptable classification accuracy. In the proposed method, a small amount of data is extracted from the under-study network as the real data. Then, the third harmonic angle of the current is extracted from the real data by an adaptive linear neuron (ADALINE) as an effective feature. The CGAN is performed to produce a large amount of pseudo data. At last, the fault data is separated from other transient network events via the CNN classifier. Five different scenarios are used to evaluate the proposed method on a 13-bus IEEE network. The simulation results show that the Precision and Recall of distinguishing HIFs from other transient events is greater than 98% in all the scenarios. These results verify that the proposed scheme is very accurate despite the low dependency on input training data. © 2022 Elsevier B.V.
Publication Date: 2007
Scientia Iranica (23453605)14(6)pp. 631-640
In this paper, reinforcement learning is used in order to model the reputation of buying and selling agents. Two important factors, quality and price, are considered in the proposed model. Each selling agent learns to evaluate the reputation of buying agents, based on their profits for that seller and uses this reputation to dedicate a discount for reputable buying agents. Also, selling agents learn to maximize their expected profits by using reinforcement learning to adjust the quality and price of the products, in order to satisfy the buying agents' preferences. In contrast, buying agents evaluate the reputation of selling agents based on two different factors: Reputation based on quality and price. Therefore, buying agents avoid interacting with disreputable selling agents. In addition, the fact that buying agents can have different priorities on the quality and price of their goods is taken into account. The proposed model has been implemented with Aglet and tested in a large-sized marketplace. The results show that selling/ buying agents that use the proposed algorithms in this paper obtain more satisfaction than the other selling/buying agents. © Sharif University of Technology, December 2007.
Publication Date: 2018
Briefings in Bioinformatics (1467-5463)19(5)pp. 878-892
Experimental drug development is time-consuming, expensive and limited to a relatively small number of targets. However, recent studies show that repositioning of existing drugs can function more efficiently than de novo experimental drug development to minimize costs and risks. Previous studies have proven that network analysis is a versatile platform for this purpose, as the biological networks are used to model interactions between many different biological concepts. The present study is an attempt to review network-based methods in predicting drug targets for drug repositioning. For each method, the preferred type of data set is described, and their advantages and limitations are discussed. For each method, we seek to provide a brief description, as well as an evaluation based on its performance metrics.We conclude that integrating distinct and complementary data should be used because each type of data set reveals a unique aspect of information about an organism. We also suggest that applying a standard set of evaluation metrics and data sets would be essential in this fast-growing research domain.
Publication Date: 2023
Journal Of Optimization In Industrial Engineering (24233935)16(2)pp. 257-273
Agent-based modelling and simulation (ABMS) is one of the topics which has been extensively studied by researchers in the field of marketing and consumer behaviour. However, no such analysis has been conducted on using Agent-based modelling and simulation in marketing and consumer behaviour. An extensive bibliometric analysis, as well as a thorough visualization and science mapping, was carried out in this field from 1995 to 2022, in response to capturing recent ABMS development in this field. A total of 1210 documents from the WOS and Scopus databases were analyse d using bibliometrix R-Tool and VOS viewer. The results showed the 20 documents with the most citations were in the area of energy consumption (55%) and innovation diffusion behaviour (20%). The USA has the most publications in this field, with the production of 188 documents. The “EXPERT SYSTEMS WITH APPLICATIONS” is a productive journal publishing in this field. Generally, the major journals that publish research on the use of ABM in marketing and consumer behaviour are multidisciplinary or interdisciplinary. 6 clusters were identified based on the analysis of the most frequent key-words: Cluster 1 (multi-agent systems and consumer behaviour), Cluster 2 (agent-based simulation and SCM), Cluster 3 (ABM and energy consumption), Cluster 4 (AMB and innovation diffusion), Cluster 5 (complex system and Simulation) and Cluster 6 (ABM and TAM). Prediction is one of the goals that has attracted the most attention of ABMS researchers among many goals such as optimization, description, self-organization, and adaptability, and there are many recent works in this field. These results show that many topics that were of interest in the past, such as the ontology of ABMS, are no longer of much interest to researchers, and the attention of researchers has been directed toward issues such as the diffusion of innovation, energy consumption, and pricing in recent years. This topic can determine the appropriate approach for other researchers to research in this field. © 2023 Qazvin Islamic Azad University. All rights reserved.
Publication Date: 2021
Journal of Biomedical Informatics (15320480)115
One of the effective missions of biology and medical science is to find disease-related genes. Recent research uses gene/protein networks to find such genes. Due to false positive interactions in these networks, the results often are not accurate and reliable. Integrating multiple gene/protein networks could overcome this drawback, causing a network with fewer false positive interactions. The integration method plays a crucial role in the quality of the constructed network. In this paper, we integrate several sources to build a reliable heterogeneous network, i.e., a network that includes nodes of different types. Due to the different gene/protein sources, four gene-gene similarity networks are constructed first and integrated by applying the type-II fuzzy voter scheme. The resulting gene-gene network is linked to a disease-disease similarity network (as the outcome of integrating four sources) through a two-part disease-gene network. We propose a novel algorithm, namely random walk with restart on the heterogeneous network method with fuzzy fusion (RWRHN-FF). Through running RWRHN-FF over the heterogeneous network, disease-related genes are determined. Experimental results using the leave-one-out cross-validation indicate that RWRHN-FF outperforms existing methods. The proposed algorithm can be applied to find new genes for prostate, breast, gastric, and colon cancers. Since the RWRHN-FF algorithm converges slowly on large heterogeneous networks, we propose a parallel implementation of the RWRHN-FF algorithm on the Apache Spark platform for high-throughput and reliable network inference. Experiments run on heterogeneous networks of different sizes indicate faster convergence compared to other non-distributed modes of implementation. © 2021 Elsevier Inc.
Publication Date: 2022
Knowledge and Information Systems (02191377)64(12)pp. 3293-3324
Patient similarity assessment, which identifies patients similar to a given patient, is a fundamental component of many secondary uses of medical data. The assessment can be performed using electronic medical records (EMRs). Patient similarity measurement requires converting heterogeneous EMRs into comparable formats to calculate distance. This study presents a new data representation method for EMRs that considers the information in clinical narratives. To address the limitations of previous approaches in handling complex parts of EMR data, an unsupervised manner is proposed for building a patient representation, which integrates unstructured and structured data extracted from patients' EMRs. We employed a tree structure to model the extracted data that capture the temporal relations of multiple medical events from EMR. We processed clinical notes to extract medical concepts using Python libraries such as MedspaCy and ScispaCy and mapped entities to the Unified Medical Language System (UMLS). To capture temporal aspects of the extracted events, we developed two new relabeling methods for the non-leaf nodes of the tree. To create an embedding vector for each patient, we traversed the tree to generate sequences that the Doc2vec algorithm would use. The comprehensive evaluation of the proposed method for patient similarity and mortality prediction tasks demonstrated that our proposed model leads to lower mean-squared error (MSE), higher precision, and normalized discounted cumulative gain (NDCG) relative to baselines. © 2022, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.
With the increasing use of mobile phones and messaging services, SMS spam has become a significant issue for users. In this paper, we propose a novel approach1 to tackle this problem by using Sine-Cosine Algorithm (SCA) and Complex Multi-Layer Perceptron (C-MLP). Specifically, we apply the SCA method to reduce the dimensionality of the feature space and CMLP to improve the performance of spam detection. Also, in this paper, we investigate the effectiveness of different classification algorithms, including Support Vector Machines, Random Forests, K-nearest neighbors, Naive Bayes, bagging, and voting approaches. Our experimental results show that the proposed approach achieves high accuracy and outperforms existing methods in terms of both accuracy and F-measure. The proposed approach can be helpful in designing effective SMS spam filters and improving the overall user experience.1All the code used in this research is publicly available on the first author's GitHub repository: https://github.com/seper-sw/SMS-Spam-Detection.git © 2023 IEEE.
Publication Date: 2014
Simulation Series (07359276)46(1)pp. 77-83
In this paper we propose an agent-based model approach to determining the effects of consumer choice on aggregate demand (CCAD). Our overall goal is to better understand how the availability of information, heuristic decision making, and social norms affect: 1) total aggregate demand and 2) the aggregated demand for disposable vs. more durable goods. In the preliminary model presented here, consumer agents select among baskets of goods with different combinations of quality and disposability. Consumer choices are based on individual agent preferences and subject to a discretionary income constraint. Agents may be either maximizing, which means that they choose the best basket of goods that they can afford, or satisficing, which means that they choose the first affordable basket of goods that they can find with utility greater than their satisfaction threshold. When run at different price levels, the resulting models can be used to generate aggregated demand curves for each group of consumers. We also demonstrate that, satisficers buy more than maximizers overall. Further analysis shows that this is because maximizers focus their trading on more durable products to gain the highest utility, however satisficers purchase more disposable products because they shop for convenience rather than utility maximization.
Roozmand, O.,
Ghasem-aghaee, N.,
Hofstede, G.J.,
Nematbakhsh, M.A.,
Baraani, A.,
Verwaart, T. Publication Date: 2011
Knowledge-Based Systems (0950-7051)24(7)pp. 1075-1095
Simulating consumer decision making processes involves different disciplines such as: sociology, social psychology, marketing, and computer science. In this paper, we propose an agent-based conceptual and computational model of consumer decision-making based on culture, personality and human needs. It serves as a model for individual behavior in models that investigate system-level resulting behavior. Theoretical concepts operationalized in the model are the Power Distance dimension of Hofstede's model of national culture; Extroversion, Agreeableness and Openness of Costa and McCrae's five-factor model of personality, and social status and social responsibility needs. These factors are used to formulate the utility function, process and update the agent state, need recognition and action estimation modules of the consumer decision process. The model was validated against data on culture, personality, wealth and car purchasing from eleven European countries. It produces believable results for the differences of consumer purchasing across eleven European countries. © 2011 Elsevier B.V. All rights reserved.
Osinga, S.A.,
Kramer, M.R.,
Hofstede, G.J.,
Roozmand, O.,
Beulens, A.J. Publication Date: 2010
Lecture Notes in Economics and Mathematical Systems (00758442)645pp. 177-188
This paper investigates the effect of a selected top-down measure (whatif scenario) on actual agent behaviour and total system behaviour by means of an agent-based simulation model, when agents' behaviour cannot fully be managed because the agents are autonomous. The Chinese pork sector serves as case. A multilevel perspective is adopted: the top-down information management measures for improving pork quality, the variation in individual farmer behaviour, and the interaction structures with supply chain partners, governmental representatives and peer farmers. To improve quality, farmers need information, which they can obtain from peers, suppliers and government. Satisfaction or dissatisfaction with their personal situation initiates change of behaviour. Aspects of personality and culture affect the agents' evaluations, decisions and actions. Results indicate that both incentive (demand) and the possibility to move (quality level within reach) on farmer level are requirements for an increase of total system quality. A more informative governmental representative enhances this effect. © Springer-Verlag Berlin Heidelberg 2010.
Publication Date: 2024
Alexandria Engineering Journal (11100168)102pp. 327-338
Publication Date: 2007
Journal of Theoretical and Applied Electronic Commerce Research (07181876)2(1)pp. 1-17
In this paper, we propose a market model which Is based on reputation and reinforcement learning algorithms for buying and selling agents. Three important factors: quality, price and delivery-time are considered in the model. We take into account the fact that buying agents can have different priorities on quality, price and delivery-time of their goods and selling agents adjust their bids according to buying agents preferences. Also we have assumed that multiple selling agents may offer the same goods with different qualities, prices and delivery-times. In our model, selling agents learn to maximize their expected profits by using reinforcement learning to adjust product quality, price and delivery-time. Also each selling agent models the reputation of buying agents based on their profits for that seller and uses this reputation to consider discount for reputable buying agents. Buying agents learn to model the reputation of selling agents based on different features of goods: reputation on quality, reputation on price and reputation on delivery-time to avoid interaction with disreputable selling agents. The model has been implemented with Aglet and tested in a large-sized marketplace. The results show that selling/buying agents that model the reputation of buying/selling agents obtain more satisfaction rather than selling/buying agents who only use the reinforcement learning. © 2007 Universidad de Talca.
Publication Date: 2011
Scientia Iranica (23453605)18(6)pp. 1460-1468
Self-Organizing Maps (SOMs) are among the most well-known, unsupervised neural network approaches to clustering, which are very efficient in handling large and high dimensional datasets. The original Particle Swarm Optimization (PSO) is another algorithm discovered through simplified social model simulation, which is effective in nonlinear optimization problems and easy to implement. In the present study, we combine these two methods and introduce a new method for anomaly detection. A discussion about our method is presented, its results are compared with some other methods and its advantages over them are demonstrated. In order to apply our method, we also performed a case study on forest fire detection. Our algorithm was shown to be simple and to function better than previous ones. We can apply it to different domains of anomaly detection. In fact, we observed our method to be a generic algorithm for anomaly detection that may need few changes for implementation in different domains. © 2012 Sharif University of Technology. Production and hosting by Elsevier B.V. All rights reserved.
Publication Date: 2023
Pattern Analysis and Applications (1433755X)26(4)pp. 1715-1727
Clinical notes contain valuable patient information. These notes are written by health care providers with various scientific levels and writing styles. It might be helpful for clinicians and researchers to understand what information is essential when dealing with extensive electronic medical records. Entities recognizing them and mapping them to standard terminologies is crucial to reducing ambiguity in processing clinical notes. Although named entity recognition and entity linking are critical steps in clinical natural language processing, they can produce repetitive and low-value concepts. On the other hand, all parts of a clinical text do not share the same importance or content in predicting the patient's condition. As a result, it is necessary to identify the section in which each content item is recorded and critical concepts to extract meaning from clinical texts. In this study, these challenges have been addressed by using clinical natural language processing techniques. In addition, a set of unsupervised essential phrase extraction methods has been verified and evaluated to identify key concepts. Considering that most clinical concepts are in the form of multi-word expressions and their accurate identification requires the user to specify an n-gram range, we have proposed a shortcut method to preserve the structure of the term based on TF-IDF (Term Frequency Inverse Document Frequency). To evaluate, we have designed two types of downstream tasks (multiple and binary classification) using the capabilities of transformer-based models. The results show the proposed method's superiority in combination with the SciBERT model. Also, they offer an insight into the efficacy of general methods for extracting essential phrases from clinical notes. © 2023, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.
Publication Date: 2019
Expert Systems with Applications (0957-4174)119pp. 476-490
Influence maximization is an important issue in social network analysis domain which concerns finding the most influential nodes. Determining the influential nodes is made with respect to information diffusion models. Most of the existing models only contain trust relationships while distrust exist in social networks as well. There exist some drawbacks in limited studies where distrust relationship is involved. The most outstanding drawback is the lack of assessment on the validity of the schemes presented on how influence propagates through distrust relationships in comparison with real word propagation in social networks. In this paper, two schemes are proposed, where based on each, some new models are proposed in two classes: cascade-based and threshold-based. All models of concern here are evaluated in comparison with the benchmark models through two real data sets, the Epinions and Bitcoin OTC. Results obtained indicate the superiority of one of the proposed schemes: when a distrusted user performs an action or adopts an opinion, the target users may tend not to do it. © 2018 Elsevier Ltd
Deffuant, G.,
Roozmand, O.,
Huet, S.,
Khamzina, K.,
Nugier, A.,
Guimond, S. Publication Date: 2023
IEEE Transactions on Computational Social Systems (2329924X)10(3)pp. 922-933
In two studies about farming practices, the respondents who are particularly favorable to organic farming tend to have a higher intention to convert their farm to organic when they perceive other farmers as not very favorable to this practice. This intention can be considered as anticonformist, as it is in opposition to the general view of others. This article hypothesizes that this phenomenon can be explained by some biases on the perceptions of attitudes. It proposes an agent-based model which computes an intention based on the theory of reasoned action (TRA) and assumes some biases in the perception of others' attitudes according to the social judgment theory. It investigates the conditions on the model parameter values for which the simulations reproduce the features observed in the studies. The results show that perceptual biases are a possible explanation of anticonformist intentions. © 2014 IEEE.
Publication Date: 2025
Journal of Supercomputing (15730484)81(5)
The importance of a node, known as centrality, can be defined and measured in various ways. The main challenge of these measurements is their extension to multilayer networks. In multilayer networks, the influence of inter-layer edges compared to intra-layer edges must be considered when calculating centrality measures. Here, the primary purpose is to provide a multilayer network-specific framework to measure the importance of nodes, with special consideration for inter-layer edges and intra-layer ones. First, we considered the different centrality measures offered for multilayer networks, as well as the associated tools and packages. Next, we implemented some more informative measures specific to multilayer networks. The functionality of implemented metrics is provided for some real networks using Python. We assessed these metrics as ranking criteria and then contrasted the ranking results using three methods: intersection similarity, rank differences, and Kendall’s tau. The findings demonstrated that incorporating information from various layers enhances the effectiveness of the criteria. The final product is a publicly available Python package called MultiNetPy, available at https://github.com/Multinetpy/Multinetpy. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025.
Publication Date: 2025
Data in Brief (23523409)61
Circular RNAs (circRNAs) are a type of RNAs that play crucial roles in various biological processes. Their outstanding properties such as tissue-specific expression and high resistance to exonuclease degradation make them attractive for research. However, a comprehensive analysis tool for analyzing circRNA data is still required. Here, we present CircPac, a newly developed web-based toolset that searches databases like circBase, circBank, and circRNADisease and organizes data to provide and visualize circRNAs information. Our toolset was created using the Python programming language and its libraries, such as pandas, seaborn, and the Django framework. CircPac enables users to unify the circRNA IDs and subsequently perform various bioinformatic analyses. These analyses include retrieving basic circRNA information, identifying target miRNAs, and analyzing circRNA expression changes in various diseases. Additionally, this toolset generates ready-to-publish figures of circRNA-miRNA interactions and circRNAs expression changes in diseases. CircPac is freely accessible (at https://www.circpac.ir) and offers a user-friendly platform for biologists to efficiently conduct and visualize circRNA data analyses in an appropriate format. © 2025 The Authors
Publication Date: 2024
Iranian Conference on Electrical Engineering, ICEE (26429527)(2024)
High Impedance Arcing Faults (HIAF) have always been considered an influential factor in the protection of electric power distribution networks (EPDNs). Characteristics such as low current levels in these faults causes the malfunction of conventional protection devices because of incorrect detection. Therefore, new methods should be provided that are able to detect the HIAF from other events in the EPDN based on these characteristics. Most of the previous fault detection techniques are dependent on a massive volume of training data to detect and classify the faults and other events, requiring a lot of time for data extraction. Furthermore, in some cases accessibility to these data is too difficult and sometimes impossible. Therefore, this paper proposes a novel protection technique based on a deep-learning algorithm to detect and classify the HIAF from other events, and also to significantly reduce the dependence on a large amount of training data. The proposed technique uses a small amount of data to extend the knowledge of pretrained SqueezeNet architecture to HIAF detection and classification problems, thereby reducing the dependence of the method on a large amount of training data. The simulation results in the presence of renewable energy sources on the modified IEEE 13-bus and 34-bus EPDNs indicate the high accuracy of the proposed technique in categorizing different network events. © 2024 IEEE.
Publication Date: 2020
Expert Systems with Applications (0957-4174)159
Background and objective: Heterogeneous complex networks are large graphs consisting of different types of nodes and edges. The knowledge extraction from these networks is complicated. Moreover, the scale of these networks is steadily increasing. Thus, scalable methods are required. Methods: In this paper, two distributed label propagation algorithms for heterogeneous networks, namely DHLP-1 and DHLP-2 have been introduced. Biological networks are one type of the heterogeneous complex networks. As a case study, we have measured the efficiency of our proposed DHLP-1 and DHLP-2 algorithms on a biological network consisting of drugs, diseases, and targets. The subject we have studied in this network is drug repositioning but our algorithms can be used as general methods for heterogeneous networks other than the biological network. Results: We compared the proposed algorithms with similar non-distributed versions of them namely MINProp and Heter-LP. The experiments revealed the good performance of the algorithms in terms of running time and accuracy. © 2020 Elsevier Ltd
Publication Date: 2025
Iranian Conference on Electrical Engineering, ICEE (26429527)pp. 56-61
Accurate differentiation between Inrush Current (IC) and internal faults (IFs) is crucial for power transformer differential protection. During transformer energization and IC generation, transformer differential protection may misinterpret IC as an internal fault, leading to a trip command being issued to the breakers. This study proposes an innovative machine learning-based approach named the tree-based pipeline optimization tool (TPOT) to enhance the F-score and efficiency of IC detection in relation to IFs in power transformers. TPOT performs in-depth data analysis and extracts significant features that influence the distinction between IC and transformer internal faults. As a model optimizer, TPOT fine-tunes models by adjusting parameters and structures. Consequently, this approach enables differentiation between IC and IFs in power transformers with high F -score and continuous improvements in detection capability. Simulation results on a real 160 MVA, 230/63 kV transformer in the MATLAB and Python software environments demonstrate the effectiveness of the proposed protection scheme in classifying transformer IC from IFs with an F1-score of 92%. © 2025 IEEE.
Publication Date: 2023
Electric Power Systems Research (03787796)219
Power transformer protection performs an essential role in power systems, ensuring a reliable power supply to the customers. One of the main challenges in differential protection of the transformers is to correctly discriminate inrush currents from internal faults and prevent the maloperation of the differential relay. In this regard, a novel differential protection method is proposed, which decomposes the differential current signal to multiple energy levels through the multi-resolution analysis (MRA) and selects the most useful feature to feed to the bidirectional gated recurrent unit (BIGRU) to classify the events. The use of the BIGRU results in the high accuracy and low implementation complexity of the proposed approach. Various simulations carried out on a 70 MVA transformer demonstrate that the proposed approach has an accuracy of 99.70% in discriminating inrush currents from internal faults in less than one-eighth of the power cycle. © 2023 Elsevier B.V.
Publication Date: 2022
IEEE Access (21693536)10pp. 120592-120605
Despite significant advances and innovations in deep network-based vehicle detection methods, finding a balance between detector accuracy and speed remains a significant challenge. This study aims to present an algorithm that can manage the speed and accuracy of the detector in real-time vehicle detection while increasing detector speed with accuracy comparable to high-speed detectors. To this end, the Fast-Yolo-Rec algorithm is proposed. The proposed method includes a new Yolo-based detection network and LSTM-based position prediction networks. The proposed semantic attention mechanism in the spatial semantic attention module (SSAM) significantly impacts accuracy and speed on par with the most recent fast detectors. Recurrent position prediction networks, on the other hand, improve the detection speed by estimating the current vehicle position using vehicle position history. The vehicle trajectories are classified, and the LSTM network for the specified trajectory predicts the vehicle positions. The Fast-Yolo-Rec algorithm not only determines the position of the vehicle faster than high-speed detectors but also allows for the speed control of the detection network with acceptable accuracy. The evaluation results on a large Highway dataset show that the proposed scheme outperforms the baseline methods. © 2013 IEEE.
