Pub Date : 2026-02-01Epub Date: 2025-12-18DOI: 10.1016/j.compeleceng.2025.110909
Chinmay Bera , Rajib Mandal , Amitesh Kumar
Accurate estimation of the State-of-Charge (SoC) in lithium-ion batteries (LIBs) is essential for optimizing performance, ensuring safety, and prolonging battery life in Battery Management Systems (BMS) for Electric Vehicles (EVs). While Long Short-Term Memory (LSTM) networks have shown significant promise for SoC estimation, they often rely on manual hyperparameter tuning, leading to inconsistent accuracy and reduced adaptability. To overcome these limitations, this study introduces a robust, noise-resilient, and adaptive deep learning framework—MRFOSA-LSTM, that combines Manta Ray Foraging Optimization (MRFO) with Simulated Annealing (SA) to automate LSTM hyperparameter tuning. The hybrid MRFOSA enhances convergence and avoids local optima, while the addition of controlled noise during training improves the model’s robustness to external interference. The proposed method is rigorously analyzed and validated using multiple real-world driving cycles and evaluated across a wide range of initial SoC levels. Comparative analysis against baseline methods, including EKF, Particle swarm optimization (PSO) based LSTM, Genetic algorithm (GA) based LSTM, MRFO-LSTM, Transformer and Bi-LSTM methods, confirms the superior performance of MRFOSA-LSTM, achieving a Mean Absolute Error (MAE) of 0.25% and Root Mean Square Error (RMSE) of 0.36%. This framework offers a highly accurate and resilient solution for real-time SoC estimation in LIBs.
{"title":"A noise-resilient adaptive deep learning framework for accurate state-of-charge prediction in lithium-ion batteries for electric vehicles","authors":"Chinmay Bera , Rajib Mandal , Amitesh Kumar","doi":"10.1016/j.compeleceng.2025.110909","DOIUrl":"10.1016/j.compeleceng.2025.110909","url":null,"abstract":"<div><div>Accurate estimation of the State-of-Charge (SoC) in lithium-ion batteries (LIBs) is essential for optimizing performance, ensuring safety, and prolonging battery life in Battery Management Systems (BMS) for Electric Vehicles (EVs). While Long Short-Term Memory (LSTM) networks have shown significant promise for SoC estimation, they often rely on manual hyperparameter tuning, leading to inconsistent accuracy and reduced adaptability. To overcome these limitations, this study introduces a robust, noise-resilient, and adaptive deep learning framework—MRFOSA-LSTM, that combines Manta Ray Foraging Optimization (MRFO) with Simulated Annealing (SA) to automate LSTM hyperparameter tuning. The hybrid MRFOSA enhances convergence and avoids local optima, while the addition of controlled noise during training improves the model’s robustness to external interference. The proposed method is rigorously analyzed and validated using multiple real-world driving cycles and evaluated across a wide range of initial SoC levels. Comparative analysis against baseline methods, including EKF, Particle swarm optimization (PSO) based LSTM, Genetic algorithm (GA) based LSTM, MRFO-LSTM, Transformer and Bi-LSTM methods, confirms the superior performance of MRFOSA-LSTM, achieving a Mean Absolute Error (MAE) of 0.25% and Root Mean Square Error (RMSE) of 0.36%. This framework offers a highly accurate and resilient solution for real-time SoC estimation in LIBs.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110909"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789757","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-16DOI: 10.1016/j.compeleceng.2025.110912
G.K. Jaiswal, U. Nangia, N.K. Jain
This research introduces a novel hybrid algorithm that combines opposition-based strategies with Differential Evolution and the Giza Pyramids Construction algorithm to address the deterministic and stochastic Optimal Reactive Power Dispatch (ORPD) problem in power systems. This novel algorithm is initially evaluated on thirteen benchmark functions, including unimodal and multimodal functions. It is then applied to single-objective deterministic ORPD problems in IEEE 30-bus and IEEE 57-bus systems, and further extended to a stochastic ORPD problem in a modified IEEE 30-bus system. In the stochastic ORPD problem, the uncertainties in load demand, wind speed, solar irradiation, and small-hydro inflows are considered. These uncertainties account for the continuous fluctuations and intrinsic intermittency of solar irradiation, wind speed, water flow rate and demand fluctuation. To demonstrate the robustness of the proposed hybrid algorithm, a comparative analysis is conducted against the recently introduced Giza Pyramids Construction Algorithm (GPC), Honey Badger Algorithm (HBA), and COOT Algorithm (COOT). For the deterministic ORPD problem, the proposed method achieves the highest savings among all four methods for , VD and VSI that are 21.75%, 92.54%, and 32.95% for the IEEE 30-bus system and 18.12%, 61.51% and 38.42% for the IEEE 57-bus system, respectively. For the stochastic ORPD problem, the proposed method obtained the expected sum of , VD and VSI as 3.8425 MW, 0.0592 p.u., and 0.0771 p.u., respectively.
{"title":"A novel hybrid optimization approach for stochastic reactive power dispatch in hybrid energy systems","authors":"G.K. Jaiswal, U. Nangia, N.K. Jain","doi":"10.1016/j.compeleceng.2025.110912","DOIUrl":"10.1016/j.compeleceng.2025.110912","url":null,"abstract":"<div><div>This research introduces a novel hybrid algorithm that combines opposition-based strategies with Differential Evolution and the Giza Pyramids Construction algorithm to address the deterministic and stochastic Optimal Reactive Power Dispatch (ORPD) problem in power systems. This novel algorithm is initially evaluated on thirteen benchmark functions, including unimodal and multimodal functions. It is then applied to single-objective deterministic ORPD problems in IEEE 30-bus and IEEE 57-bus systems, and further extended to a stochastic ORPD problem in a modified IEEE 30-bus system. In the stochastic ORPD problem, the uncertainties in load demand, wind speed, solar irradiation, and small-hydro inflows are considered. These uncertainties account for the continuous fluctuations and intrinsic intermittency of solar irradiation, wind speed, water flow rate and demand fluctuation. To demonstrate the robustness of the proposed hybrid algorithm, a comparative analysis is conducted against the recently introduced Giza Pyramids Construction Algorithm (GPC), Honey Badger Algorithm (HBA), and COOT Algorithm (COOT). For the deterministic ORPD problem, the proposed method achieves the highest savings among all four methods for <span><math><msub><mrow><mi>P</mi></mrow><mrow><mi>L</mi><mi>o</mi><mi>s</mi><mi>s</mi></mrow></msub></math></span>, VD and VSI that are 21.75%, 92.54%, and 32.95% for the IEEE 30-bus system and 18.12%, 61.51% and 38.42% for the IEEE 57-bus system, respectively. For the stochastic ORPD problem, the proposed method obtained the expected sum of <span><math><msub><mrow><mi>P</mi></mrow><mrow><mi>L</mi><mi>o</mi><mi>s</mi><mi>s</mi></mrow></msub></math></span>, VD and VSI as 3.8425 MW, 0.0592 p.u., and 0.0771 p.u., respectively.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110912"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
These days, gastrointestinal disorders are a major health concern, and serious consequences can be avoided with early detection. An effective tool for the examiner to make an accurate diagnosis of the disease is a computer-aided diagnosis (CAD) system. However, the developed artificial intelligence (AI) algorithm determines the power consumption and latency of the CAD system. The AI algorithm needs to be optimized for edge devices for real-time implementation. In the proposed LDFGastro-Net, we have developed a lightweight hybrid convolutional neural network (CNN) model for the classification of gastrointestinal disorders. The initial layer is derived from the MobileNet-V2 pre-trained model for the extraction of low-level features, as the proposed model is intended for Field Programmable Gate Array (FPGA) deployment and must be lightweight. Next, a dense structure of depth-wise separable layers forms the middle section of the proposed framework. The dense connection has the advantage of feature reuse with the extraction of essential spatial features along with low-level features. The depthwise separable and feature fusion, which help in class-specific features and preservation of low level features, are included in the final layers. The proposed model’s performance has been demonstrated through Grad-CAM visualizations, highlighting its ability to classify gastrointestinal disorders better. With an accuracy of 98.2%, the proposed model outperforms the existing custom CNN model and several state-of-the-art pretrained architectures.
{"title":"LDFGastro-Net: Lite-DenseFuse Network for gastrointestinal disorders classification towards hardware deployment","authors":"Debaraj Rana , Bunil Kumar Balabantaray , Rajashree Nayak , Rangababu Peesapati","doi":"10.1016/j.compeleceng.2025.110852","DOIUrl":"10.1016/j.compeleceng.2025.110852","url":null,"abstract":"<div><div>These days, gastrointestinal disorders are a major health concern, and serious consequences can be avoided with early detection. An effective tool for the examiner to make an accurate diagnosis of the disease is a computer-aided diagnosis (CAD) system. However, the developed artificial intelligence (AI) algorithm determines the power consumption and latency of the CAD system. The AI algorithm needs to be optimized for edge devices for real-time implementation. In the proposed LDFGastro-Net, we have developed a lightweight hybrid convolutional neural network (CNN) model for the classification of gastrointestinal disorders. The initial layer is derived from the MobileNet-V2 pre-trained model for the extraction of low-level features, as the proposed model is intended for Field Programmable Gate Array (FPGA) deployment and must be lightweight. Next, a dense structure of depth-wise separable layers forms the middle section of the proposed framework. The dense connection has the advantage of feature reuse with the extraction of essential spatial features along with low-level features. The depthwise separable and feature fusion, which help in class-specific features and preservation of low level features, are included in the final layers. The proposed model’s performance has been demonstrated through Grad-CAM visualizations, highlighting its ability to classify gastrointestinal disorders better. With an accuracy of 98.2%, the proposed model outperforms the existing custom CNN model and several state-of-the-art pretrained architectures.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110852"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145572090","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Automatic speech recognition (ASR) systems powered by deep neural networks require substantial labeled speech data to achieve high performance. However, acquiring high-quality speech-text pairs remains costly and time-consuming, particularly for low-resource languages where data scarcity and limited diversity pose significant challenges. To mitigate these challenges, data augmentation (DA) techniques create synthetic training samples from existing data, effectively improving model robustness and performance. Given the critical role of data augmentation in advancing ASR systems, this paper presents the first comprehensive review of DA techniques for ASR, addressing a significant gap in the literature. Application of data augmentation to ASR systems introduces unique complexities stemming from the multifaceted nature of speech signals. We examine these speech-specific constraints to equip readers with the necessary background information for current approaches and future research directions. We propose a structured taxonomy of existing ASR data augmentation approaches, categorized along five key dimensions: Data Creation Methodology, Augmentation Modality, Automation Approach, Training Paradigm, and Application Time. Our review spans signal-based to advanced deep learning-based approaches, providing a systematic analysis of DA methods for ASR. By analyzing strengths and limitations of each method in-depth, we guide researchers in selecting appropriate techniques for their practical requirements. Furthermore, we discuss key challenges and promising research directions, including evaluation methodologies, automatic DA strategies, multi-variant augmentation, leveraging large language models, and theoretical understanding of speech augmentation. Our review can serve as a reference for providing in-depth knowledge of existing ASR data augmentation methods, identifying key challenges, and paving the way for future research.
{"title":"Data augmentation techniques for Automatic Speech Recognition: Taxonomy, method analysis, challenges, and future research directions","authors":"Maryam Asadolahzade Kermanshahi , Ahmad Akbari Azirani , Babak Nasersharif , Seyed Jahanshah Kabudian","doi":"10.1016/j.compeleceng.2025.110851","DOIUrl":"10.1016/j.compeleceng.2025.110851","url":null,"abstract":"<div><div>Automatic speech recognition (ASR) systems powered by deep neural networks require substantial labeled speech data to achieve high performance. However, acquiring high-quality speech-text pairs remains costly and time-consuming, particularly for low-resource languages where data scarcity and limited diversity pose significant challenges. To mitigate these challenges, data augmentation (DA) techniques create synthetic training samples from existing data, effectively improving model robustness and performance. Given the critical role of data augmentation in advancing ASR systems, this paper presents the first comprehensive review of DA techniques for ASR, addressing a significant gap in the literature. Application of data augmentation to ASR systems introduces unique complexities stemming from the multifaceted nature of speech signals. We examine these speech-specific constraints to equip readers with the necessary background information for current approaches and future research directions. We propose a structured taxonomy of existing ASR data augmentation approaches, categorized along five key dimensions: Data Creation Methodology, Augmentation Modality, Automation Approach, Training Paradigm, and Application Time. Our review spans signal-based to advanced deep learning-based approaches, providing a systematic analysis of DA methods for ASR. By analyzing strengths and limitations of each method in-depth, we guide researchers in selecting appropriate techniques for their practical requirements. Furthermore, we discuss key challenges and promising research directions, including evaluation methodologies, automatic DA strategies, multi-variant augmentation, leveraging large language models, and theoretical understanding of speech augmentation. Our review can serve as a reference for providing in-depth knowledge of existing ASR data augmentation methods, identifying key challenges, and paving the way for future research.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110851"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145618605","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-10DOI: 10.1016/j.compeleceng.2025.110904
Muhammad Ali Iqbal, Joon-Min Gil, Soo Kyun Kim
Open-world object detection (OWOD) extends traditional detection to dynamic environments where models must both recognize an expanding set of known categories and discover previously unseen objects. Existing OWOD methods mine unknowns by thresholding objectness scores from class-agnostic proposals, but this induces a label bias: novel objects that diverge from known-category appearances are suppressed as background, while complex backgrounds may spuriously trigger unknown detections. To overcome these limitations, the proposed REConstruction-Error Density + Contrastive Quality Architecture (RcCOnQA) decouples pseudo-label generation from objectness via reconstruction-error density (RED) modeling. A lightweight Transformer autoencoder reconstructs frozen backbone + Feature Pyramid Network (FPN) features and produces per-anchor residual maps; a compact density head then converts normalized residuals into continuous ’unknownness’ scores. These scores guide a self-training detector enhanced with an object-localization network (OLN) branch and a contrastive Quality Head, enabling precise pseudo-label refinement and task-incremental learning without catastrophic forgetting. Experiments on the Open-World Object Detection Benchmarks (OWODB), including superclass-separated split (S-OWODB) and superclass-mixed split (M-OWODB) demonstrate substantial improvements in unknown recall while preserving known-class mean Average Precision (mAP) compared to state-of-the-art OWOD approaches. Cross-dataset evaluations on Large Vocabulary Instance Segmentation (LVIS) and Objects365, along with semantic-relatedness studies, confirm robust generalization to truly out-of-distribution objects.
{"title":"An incremental out-of-distribution learning framework for robust open-world object detection","authors":"Muhammad Ali Iqbal, Joon-Min Gil, Soo Kyun Kim","doi":"10.1016/j.compeleceng.2025.110904","DOIUrl":"10.1016/j.compeleceng.2025.110904","url":null,"abstract":"<div><div>Open-world object detection (OWOD) extends traditional detection to dynamic environments where models must both recognize an expanding set of known categories and discover previously unseen objects. Existing OWOD methods mine unknowns by thresholding objectness scores from class-agnostic proposals, but this induces a label bias: novel objects that diverge from known-category appearances are suppressed as background, while complex backgrounds may spuriously trigger unknown detections. To overcome these limitations, the proposed REConstruction-Error Density + Contrastive Quality Architecture (RcCOnQA) decouples pseudo-label generation from objectness via reconstruction-error density (RED) modeling. A lightweight Transformer autoencoder reconstructs frozen backbone + Feature Pyramid Network (FPN) features and produces per-anchor residual maps; a compact density head then converts normalized residuals into continuous ’unknownness’ scores. These scores guide a self-training detector enhanced with an object-localization network (OLN) branch and a contrastive Quality Head, enabling precise pseudo-label refinement and task-incremental learning without catastrophic forgetting. Experiments on the Open-World Object Detection Benchmarks (OWODB), including superclass-separated split (S-OWODB) and superclass-mixed split (M-OWODB) demonstrate substantial improvements in unknown recall while preserving known-class mean Average Precision (mAP) compared to state-of-the-art OWOD approaches. Cross-dataset evaluations on Large Vocabulary Instance Segmentation (LVIS) and Objects365, along with semantic-relatedness studies, confirm robust generalization to truly out-of-distribution objects.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110904"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145736918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-04DOI: 10.1016/j.compeleceng.2025.110867
Raseena T.P. , Balasundaram S.R. , Jitendra Kumar
Colorectal Cancer (CRC) is the second leading cause of cancer-related deaths worldwide, since early and precise classification of colorectal polyps is vital for reducing mortality. While the existing deep learning-based classification approaches are effective, most of them predominantly focus on either local texture or global semantic features. It limits their ability to fully capture the complex morphology of polyps, often resulting in increased false detection rates. To address such challenges, this study proposes a novel framework, VGG19–Graph Convolutional Network (V-GCN), which introduces inter-image relational reasoning through a graph-based feature fusion framework designed for context-aware learning. By modeling relationships among semantically similar images through a graph structure, V-GCN captures long-range spatial dependencies and global contextual patterns more effectively. This inter-image relational modeling introduces a new perspective in polyp classification by leveraging graph-based global context beyond individual image boundaries. The proposed model also integrates an efficient image enhancement technique, Bilateral Filtered-Discrete Wavelet Transform Network (BF-DWT Net), to enrich visual quality by combining bilateral filtering for noise suppression with the discrete wavelet transform for multiscale edge enhancement to preserve subtle structural details. The enhanced images are first processed by VGG19 to extract fine-grained hierarchical features, which are subsequently refined by the graph convolutional network to enhance global contextual representation and local discriminative detail. Experiments on four benchmark datasets demonstrate the superior performance of V-GCN, with accuracies of 81.88% on PolypsSet, 99.89% on CPchildA, 99.75% on CPchildB, and 98.50% on KvasirV2, highlighting the significance of combining inter-image graph reasoning with multiscale feature enhancement.
{"title":"Graph-based feature fusion network with multiscale edge-preserving techniques for polyp identification","authors":"Raseena T.P. , Balasundaram S.R. , Jitendra Kumar","doi":"10.1016/j.compeleceng.2025.110867","DOIUrl":"10.1016/j.compeleceng.2025.110867","url":null,"abstract":"<div><div>Colorectal Cancer (CRC) is the second leading cause of cancer-related deaths worldwide, since early and precise classification of colorectal polyps is vital for reducing mortality. While the existing deep learning-based classification approaches are effective, most of them predominantly focus on either local texture or global semantic features. It limits their ability to fully capture the complex morphology of polyps, often resulting in increased false detection rates. To address such challenges, this study proposes a novel framework, VGG19–Graph Convolutional Network (V-GCN), which introduces inter-image relational reasoning through a graph-based feature fusion framework designed for context-aware learning. By modeling relationships among semantically similar images through a graph structure, V-GCN captures long-range spatial dependencies and global contextual patterns more effectively. This inter-image relational modeling introduces a new perspective in polyp classification by leveraging graph-based global context beyond individual image boundaries. The proposed model also integrates an efficient image enhancement technique, Bilateral Filtered-Discrete Wavelet Transform Network (BF-DWT Net), to enrich visual quality by combining bilateral filtering for noise suppression with the discrete wavelet transform for multiscale edge enhancement to preserve subtle structural details. The enhanced images are first processed by VGG19 to extract fine-grained hierarchical features, which are subsequently refined by the graph convolutional network to enhance global contextual representation and local discriminative detail. Experiments on four benchmark datasets demonstrate the superior performance of V-GCN, with accuracies of 81.88% on PolypsSet, 99.89% on CPchildA, 99.75% on CPchildB, and 98.50% on KvasirV2, highlighting the significance of combining inter-image graph reasoning with multiscale feature enhancement.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110867"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684721","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-20DOI: 10.1016/j.compeleceng.2025.110816
Mohamed H. Hassan , Salah Kamel , Ehab Mahmoud Mohamed
Optimal Reactive Power Dispatch (ORPD) has emerged as a vital requirement for the safe, efficient, and economical operation of power networks. This study presents a leader-based enhancement to the original Wild Horse Optimizer (WHO), resulting in a more powerful algorithm referred to as LWHO. The performance of the LWHO algorithm is rigorously evaluated using 23 mathematical benchmark functions, encompassing unimodal, multimodal, and composite optimization problems.
Furthermore, both single-objective and multi-objective deterministic/stochastic ORPD formulations are examined on two standard test systems: the IEEE 30-bus and IEEE 57-bus networks. To effectively model uncertainty, a scenario-based approach is utilized, incorporating variations in load demand and RES output. Simulation results confirm that the proposed LWHO algorithm delivers highly accurate and robust solutions for ORPD under uncertainty. Statistical validation using the Wilcoxon rank-sum test confirms the significant superiority of the proposed LWHO compared to the original WHO in five out of eight single-objective cases (p < 0.05). This method offers a practical and efficient strategy for addressing the complexities introduced by RES integration, ultimately contributing to enhanced energy efficiency and more resilient power system operations.
{"title":"A leader-driven Wild Horse Optimizer for solving ORPD with integrated stochastic renewable sources","authors":"Mohamed H. Hassan , Salah Kamel , Ehab Mahmoud Mohamed","doi":"10.1016/j.compeleceng.2025.110816","DOIUrl":"10.1016/j.compeleceng.2025.110816","url":null,"abstract":"<div><div>Optimal Reactive Power Dispatch (ORPD) has emerged as a vital requirement for the safe, efficient, and economical operation of power networks. This study presents a leader-based enhancement to the original Wild Horse Optimizer (WHO), resulting in a more powerful algorithm referred to as LWHO. The performance of the LWHO algorithm is rigorously evaluated using 23 mathematical benchmark functions, encompassing unimodal, multimodal, and composite optimization problems.</div><div>Furthermore, both single-objective and multi-objective deterministic/stochastic ORPD formulations are examined on two standard test systems: the IEEE 30-bus and IEEE 57-bus networks. To effectively model uncertainty, a scenario-based approach is utilized, incorporating variations in load demand and RES output. Simulation results confirm that the proposed LWHO algorithm delivers highly accurate and robust solutions for ORPD under uncertainty. Statistical validation using the Wilcoxon rank-sum test confirms the significant superiority of the proposed LWHO compared to the original WHO in five out of eight single-objective cases (<em>p</em> < 0.05). This method offers a practical and efficient strategy for addressing the complexities introduced by RES integration, ultimately contributing to enhanced energy efficiency and more resilient power system operations.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110816"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145572088","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-20DOI: 10.1016/j.compeleceng.2025.110860
Arnab Kumar Das, Ruchira Naskar
The recent years have seen a rapid rise in deepfakes on all social platforms, of celebrities and reputed personalities, majorly aimed towards defamation and victimization of the person concerned.
Researchers have started exploring frame-based deepfake detection techniques, most of which fail with increased realism in deepfakes caused by stronger, more realistic deepfake generators. The direction of multi-modal deepfakes proves promising, with recent researches reporting improved results through adoption of multiple modes for deepfake detection.
In this work, we propose a Hybrid, Triple-Modality and Deep-Features based Deepfake Detection (HTMDF-DD) framework, which exploits three distinct modes for detection, viz., audio, text, and visual modalities. HTMDF-DD works in two stages: first, it extracts spatio-temporal information from the visual domain, and second, it tries to reconstruct this information using the auditory and text (language) domains. This process enables triple-modality interactions, based on which we successfully detect a deepfake video. The source code of our proposed HTMDF-DD framework is publicly available on the GitHub link: https://github.com/arnabdasphd/HTMDF-DD.
{"title":"HTMDF-DD: Hybrid triple modality based spatial–temporal features early fusion for deepfake detection","authors":"Arnab Kumar Das, Ruchira Naskar","doi":"10.1016/j.compeleceng.2025.110860","DOIUrl":"10.1016/j.compeleceng.2025.110860","url":null,"abstract":"<div><div>The recent years have seen a rapid rise in deepfakes on all social platforms, of celebrities and reputed personalities, majorly aimed towards defamation and victimization of the person concerned.</div><div>Researchers have started exploring frame-based deepfake detection techniques, most of which fail with increased realism in deepfakes caused by stronger, more realistic deepfake generators. The direction of multi-modal deepfakes proves promising, with recent researches reporting improved results through adoption of multiple modes for deepfake detection.</div><div>In this work, we propose a <u>H</u>ybrid, <u>T</u>riple-<u>M</u>odality and <u>D</u>eep-<u>F</u>eatures based <u>D</u>eepfake <u>D</u>etection (HTMDF-DD) framework, which exploits three distinct modes for detection, viz., audio, text, and visual modalities. HTMDF-DD works in two stages: first, it extracts spatio-temporal information from the visual domain, and second, it tries to reconstruct this information using the auditory and text (language) domains. This process enables triple-modality interactions, based on which we successfully detect a deepfake video. The source code of our proposed HTMDF-DD framework is publicly available on the GitHub link: <span><span>https://github.com/arnabdasphd/HTMDF-DD</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110860"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145572054","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-25DOI: 10.1016/j.compeleceng.2025.110857
Chiemeka L. Maxwell , Dongsheng Yu , Yang Leng
This paper presents a lightweight encryption strategy for secure communication in power and data synchronous transmission (PDST) systems, leveraging a stochastic memristor-based two-dimensional (2D) chaotic map. While PDST frameworks have advanced in modulation design and converter integration, the security of the transmitted data remains an underexplored challenge - particularly for embedded, real-time applications. To address this, we propose a coupled memristor-tent map system that generates high-entropy pseudorandom sequences through bidirectional nonlinear interactions, forming the basis of a robust key generation scheme. The resulting hyperchaotic system exhibits strong sensitivity to initial conditions and supports cryptographic operations on the DC bus of switched-mode power supplies (SMPSs). A differential quadrature phase shift keying (DQPSK) scheme is adopted for communication. We implement the encryption framework in MATLAB/Simulink and validate its performance through entropy analysis, bit error rate (BER) analysis, and full NIST randomness evaluation. A hardware-in-the-loop (HIL) setup is used to demonstrate the PDST system in real-time. Results confirm that the proposed method achieves high randomness, and strong resistance to algebraic and differential attacks, making it well-suited for secure and scalable deployment in industrial PDST applications.
{"title":"A lightweight hyperchaotic memristor-based encryption for secure power and data synchronous transmission in DC microgrids","authors":"Chiemeka L. Maxwell , Dongsheng Yu , Yang Leng","doi":"10.1016/j.compeleceng.2025.110857","DOIUrl":"10.1016/j.compeleceng.2025.110857","url":null,"abstract":"<div><div>This paper presents a lightweight encryption strategy for secure communication in power and data synchronous transmission (PDST) systems, leveraging a stochastic memristor-based two-dimensional (2D) chaotic map. While PDST frameworks have advanced in modulation design and converter integration, the security of the transmitted data remains an underexplored challenge - particularly for embedded, real-time applications. To address this, we propose a coupled memristor-tent map system that generates high-entropy pseudorandom sequences through bidirectional nonlinear interactions, forming the basis of a robust key generation scheme. The resulting hyperchaotic system exhibits strong sensitivity to initial conditions and supports cryptographic operations on the DC bus of switched-mode power supplies (SMPSs). A differential quadrature phase shift keying (DQPSK) scheme is adopted for communication. We implement the encryption framework in MATLAB/Simulink and validate its performance through entropy analysis, bit error rate (BER) analysis, and full NIST randomness evaluation. A hardware-in-the-loop (HIL) setup is used to demonstrate the PDST system in real-time. Results confirm that the proposed method achieves high randomness, and strong resistance to algebraic and differential attacks, making it well-suited for secure and scalable deployment in industrial PDST applications.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110857"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145618607","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-05DOI: 10.1016/j.compeleceng.2025.110866
Roya Aghadavoud Marnani , Michal Podpora , Aleksandra Kawala-Sterniuk , Xuyuan Tao , Petr Bilik , Radek Martinek
The continuous monitoring of biosignals plays a crucial role in identifying and addressing serious health conditions. Conventional silver/silver chloride (Ag/AgCl) electrodes, typically used for measuring biosignals, pose challenges for extended monitoring due to their tendency to dry out and cause skin irritation over time. Textile electrodes (TE) are a promising alternative that effectively overcomes the limitations of traditional electrodes. They offer enhanced comfort and usability, improving healthcare diagnostics. This review aims to provide a broad overview and critical evaluation of various materials and methods for TE production. Furthermore, technical challenges including TE shape and size, electrode skin impedance, and signal processing are discussed. Despite the advantages that TE provides, their challenges persist. These electrodes record biosignals and noises, necessitating signal-processing methods for accurate interpretation and analysis of biosignals. Moreover, the absence of conductive paste in TEs results in higher electrode skin impedance. TEs can be manufactured in various shapes, designs, and sizes. However, there is a lack of universal standards for these parameters. Ongoing research focus on developing advanced noise reduction algorithms and standards for TE production, potentially enhancing biosignal monitoring and facilitating early anomaly detection.
{"title":"Challenges in biomedical signals monitoring using textile electrodes: A review","authors":"Roya Aghadavoud Marnani , Michal Podpora , Aleksandra Kawala-Sterniuk , Xuyuan Tao , Petr Bilik , Radek Martinek","doi":"10.1016/j.compeleceng.2025.110866","DOIUrl":"10.1016/j.compeleceng.2025.110866","url":null,"abstract":"<div><div>The continuous monitoring of biosignals plays a crucial role in identifying and addressing serious health conditions. Conventional silver/silver chloride (Ag/AgCl) electrodes, typically used for measuring biosignals, pose challenges for extended monitoring due to their tendency to dry out and cause skin irritation over time. Textile electrodes (TE) are a promising alternative that effectively overcomes the limitations of traditional electrodes. They offer enhanced comfort and usability, improving healthcare diagnostics. This review aims to provide a broad overview and critical evaluation of various materials and methods for TE production. Furthermore, technical challenges including TE shape and size, electrode skin impedance, and signal processing are discussed. Despite the advantages that TE provides, their challenges persist. These electrodes record biosignals and noises, necessitating signal-processing methods for accurate interpretation and analysis of biosignals. Moreover, the absence of conductive paste in TEs results in higher electrode skin impedance. TEs can be manufactured in various shapes, designs, and sizes. However, there is a lack of universal standards for these parameters. Ongoing research focus on developing advanced noise reduction algorithms and standards for TE production, potentially enhancing biosignal monitoring and facilitating early anomaly detection.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"130 ","pages":"Article 110866"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145685508","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号