Manual grading of Ribbed Smoked Sheets (RSS) is labour-intensive and experience-dependent, which can lead to inconsistent grade assignment and reduced transparency in quality evaluation. This paper presents a low-cost machine-vision inspection system for full-spectrum RSS grading (RSS1–RSS5) that integrates standardised image acquisition and HSV-based colour representation with a lightweight convolutional neural network (CNN) classifier. Images were captured using a custom-built acquisition setup with fixed illumination and camera geometry to improve repeatability across samples. To enhance grading-relevant cue representation, RGB images were transformed into the Hue–Saturation–Value (HSV) colour space, separating chromatic information from brightness prior to classification. Using a balanced dataset of expert-labelled images, the HSV-based system achieved 91.87% classification accuracy, outperforming an RGB baseline (87.81%). Confusion-matrix analysis indicated strong performance across all five grades, with remaining errors concentrated between visually similar adjacent grades. The results demonstrate the feasibility of combining low-cost standardised illumination with HSV-based colour processing and lightweight learning for practical RSS inspection and grading.

In this research, the energy consumption models of Bitcoin, Ethereum, and Dogecoin are analyzed using Explainable Artificial Intelligence (XAI) models aided by the three stages of analysis involving Digiconomist data from 2022 to 2025: (1) exploratory data analysis for the nature of energy consumption, (2) model identification of influential variables using Random Forest models enhanced with SHAP values, and (3) an LSTM transfer learning method for predicting the energy consumption of Ethereum and Dogecoin using a model developed with Bitcoin data. The initial results show that while both assets vary largely when it comes to their normal usage level, Ethereum sees a sharp drop after the changeover from Proof-of-Work to Proof-of-Stake as a mechanism. The XAI analysis indicates that energy use is largely a consequence of past use, seasonality, and annual patterns. In addition to this, the models show a high level of accuracy for Dogecoin (R²: 88.4%, MAPE: 13.45%) and Ethereum (R²: 86.2%, MAPE: 11.47%) when it comes to predicting energy usage using the concepts of transfer learning.

Modern power systems face the dual challenges of power electronic component failures and the intermittency of renewable sources. This paper presents a planning methodology for offshore MMC-HVDC-based wind power systems, using a bottom-up approach that links component-level to system-level performance. The modelling framework enables direct evaluation of component influence on overall system availability with particular focus on wind turbines and MMC components. This level of detail is integrated with O&M cost considerations to optimise converter asset management and offshore wind capacity planning, resulting in an adequacy-constrained, O&M cost-minimised planning framework, showcased in an IEEE RTS-24 system. The analysis reveals, at the component level, a 0.86\% greater downtime-related capacity factor loss in geared versus direct-driven turbines, and a 23.96% sensitivity in the availability of a 525 kV DC transmission end to MMC redundancy and maintenance. At the system level, it highlights nonlinear and converging trends in reliability and O&M costs with offshore wind penetration, as well as the significant influence of MMC asset management on overall O&M expenditures. Ultimately, the method guides the substitution of conventional generation with offshore wind, enhancing both the robustness and cost-efficiency of renewable energy integration.

This work experimentally proposes a hybrid system that combines Radio Frequency and Visible Light Communication technologies to mitigate signal non-uniformity in indoor environments. Additionally, an experimental study on the degradation of the signal-to-noise ratio of a commercial broadcast signal at a frequency of 103.5 MHz is presented. This analysis was carried out at different levels of the ITM University Institution in Medellín, including basements 1 and 2, and focused exclusively on signal reception. The results were compared with the initial measurements taken, showing that the pure RF signal, captured conventionally, had an average signal-to-noise ratio of 4.93 dB. After implementing the hybrid system, the average signal-to-noise ratio increased significantly to 36.9 dB. These results demonstrate the effectiveness of the technological convergence proposed in this study to improve signal quality in multiple demanding application scenarios, such as the Internet of Things, Smart Cities, Indoor Positioning, Sustainable Mining, Marine Monitoring, and others.

Detection of petrol adulteration at concentration of 5 % at high sensitivity is challenging. The proposed SPR sensor shows high performance in term of sensitivity and optical energy coupling efficiency (OECE) which showed the silver thickness of 37 nm gives the sensitivity of 491.3043°/RIU higher by 1.25, 2.4, and 8.07 times of the sensitivity for the silver thickness is 40, 45 and 50 nm, respectively. For silver thickness 37 nm, the OECE is 99 % higher than silver thickness is 40, 45 50 and 55 nm with OECE of 70.64, 58.52, 48.86 and 35.51 %, respectively. This indicates that the fields were enhanced by 177.61% than when the silver thickness was 55 nm. At kerosene concentrations of 5% and 40% in petrol, the sensitivity, FOM, and DA are 330.43 °/RIU, 63.9380, and 0.1634, and 116.42, 0.2370, and 491.3043 °/RIU, respectively, demonstrating improvements of 18% and 5.61% compared to a recently published research. Mechanical scanning constrained by a motor resolution of 0.001°, utilizes a diverging beam reflected onto a CCD sensor of 3,648 pixels, paired with a 12-bit ADC converter, yielding an exceptional angular resolution of 6.694×10-8°, representing an improvement of 99.98% above mechanical scanning. Its resolution is 7×10-10 RIU.

In SPR sensor multilayer is important process in improving the performance of the sensor chip. However, during the coating process of these multilayers the coated layer is highly susceptible to roughness the effect becomes more when the size of the chip is bigger, coated which leads to surface roughness. Thus, the research focuses on quantifying the impacts of surface roughness of multilayers on the performance of the SPR sensor. The Whittaker filter was proposed to mitigate the roughness effect without changing the hardware part. The study found that – sample interface roughness has the most negative impact on SPR sensor performance. The proposed Whittaker filter showed that can recovers the distorted SPR signal caused by roughness effect at the interface the Whittaker filter recovers about 50 - 60 % of this performance loss. Surface roughness decreases detection accuracy by approximately 25 - 30%, with the filter providing about 40 - 50% recovery. The error caused by rough interfaces were reduced to the average of approximately 49% Thus in area where there is budget constraint in achieving smooth surface during coating then the Whittaker filter provides a viable compromise, recovering significant performance without physical intervention.

This study introduces a rigorous stacked ensemble framework for binary image classification, specifically targeting the Cats vs. Dogs dataset. The proposed method integrates eight transfer-learning convolutional neural networks (CNNs): Xception, VGG16, VGG19, InceptionV3, InceptionResNetV2, MobileNet, MobileNetV2, and DenseNet121 as base learners, combined via eight different meta-learners. Through extensive experimentation that included hyperparameter optimization, ablation studies, and robustness assessments (under Gaussian noise and image blur), the ensemble demonstrated superior accuracy and reliability. The optimal configuration, which utilized a Support Vector Machine (SVM) meta-learner, achieved an accuracy of 99.28%, surpassing prior benchmarks. Statistical analysis confirmed the significance of this improvement (paired t-test: t = 29.94, p < 0.0001). InceptionResNetV2 and Xception made the largest contributions to ensemble performance, as revealed by ablation studies. The model also maintained robust performance under noise (97.91% accuracy) and blur (98.43% accuracy), attesting to its suitability for practical applications such as veterinary diagnostics and pet identification. This work highlights the effectiveness of deep learning ensembles in binary image classification tasks and suggests future directions in adaptive ensemble weighting and cross-dataset validation.

Based on the CiteSpace Software tool, this article specifically analyzed the research development process and the current research status in the field of Feeder Automation from 1981 to 2024 with 1,892 papers from the core collection of Web of Science. This article adopted the methods of Knowledge Graph, from the external feature analysis, co-citation analysis, and research hotspot and frontier analysis to analyze the research status of foremost research scholars, primary cited literature, and significant research hotspots in the field of Feeder Automation in various aspects, to sort out the current research status of the field, and reveal the future research hotspots and directions in the field.

This research addresses the critical challenge of accurately discriminating inrush currents within differential relays, a key component of power transformer protection systems. Differential relays distinguish genuine internal faults from transient events, such as inrush currents and external faults. Failure to make this distinction can lead to false tripping, which, in turn, may result in significant operational disruptions and economic losses. The study employs Discrete Wavelet Transform (DWT) and Adaptive Neuro-Fuzzy Inference System (ANFIS) techniques to develop reliable and intelligent protection systems. This intelligent system distinguishes internal faults from inrush currents and external faults. This system improves traditional remedies such as harmonic blocking and restrain and wave shape recognition techniques. The paper discusses the project’s methods, MATLAB/Simulink simulations, and implications for power transformers in electrical grid stations. The results indicated that this system accurately discriminated inrush currents with a 100% success rate and 91.5% accuracy at a speed of 3.5 ms.

In order to solve the problem of safety accidents, such as the overturning of the traction walking plate and the jumping of subconductors, during the construction process of transmission line tension framing, because the construction personnel cannot fully monitor the key state quantities of the traction walking plate, a multi-parameter intelligent monitoring system for the traction walking plate based on the digital twin technology is developed. In the selection of twin data, based on the analysis of the risk points of the traction walking plate, the subconductor tension, tilt angle and spatial position of the traction walking plate are selected as the twin data. For the mapping of the twin data, the grid information model (GIM) and building information model (BIM) are used as the basis to establish a refined engineering structure model, which accurately reflects the geometry and mechanical properties of the traction slab. Finally, the connection design of twin data, hardware design, equipment data layer design and twin service type are designed in 3 parts, and the improved isolated forest algorithm is adopted as the twin service layer of the monitoring system.

To address the transient voltage stability and inter-regional power transmission fluctuations caused by the integration of distributed renewable energy into distribution networks, a distributed synchronous condenser optimal allocation method considering transient processes is proposed. First, based on the transient voltage safety ride-through standards for renewable energy, the transient voltage fluctuations and power transmission deviations of grid-connected buses under different operating conditions are analyzed, and a time-section-based transient voltage safety evaluation index is proposed. Second, an index for improving the transient voltage safety level is established, which is used to determine the installation locations of distributed synchronous condensers. On this basis, the objective function and constraints for synchronous condenser capacity allocation are formulated, and the optimization problem is solved using the Beetle Swarm Optimization (BSO) algorithm. Finally, the rationality of the distributed synchronous condenser installation locations and capacity optimization is verified through an improved IEEE 33-bus system. Simulation results demonstrate that after the optimal allocation of distributed synchronous condensers, the voltage of renewable energy buses in the distribution network meets the specified transient voltage safety ride-through standards, and the power transmission deviation is significantly reduced, validating the correctness of the conclusions and the effectiveness of the proposed method.

High-Level Synthesis (HLS) design space exploration plays a pivotal role in electronic design automation by automatically converting high-level descriptions (e.g., C/C++/SystemC) into register-transfer level (RTL) code, significantly enhancing hardware design efficiency. However, the HLS design space remains vast and complex, involving multiple factors such as computational modules, data flow, control flow, and memory allocation, making traditional heuristic searches and experience-driven methods ineffective for efficient optimization. Conventional heuristic searches and graph neural network-based learning approaches exhibit limitations when handling heterogeneous structures. Current research predominantly focuses on homogeneous graph neural networks or statistical model-based methods, neglecting the inherent heterogeneity in HLS designs. This paper presents the first heterogeneous graph neural network (HGNN)-based prediction framework that models HLS designs as heterogeneous graphs to precisely capture complex interactions among different types of computational units, data dependencies, and optimization strategies. We design a multi-layer heterogeneous graph message-passing mechanism combined with a self-supervised learning strategy to enhance prediction accuracy and generalization capability in design space exploration. Experimental results demonstrate that our model can accurately estimate latency and resource utilization for unseen designs (i.e., novel CDFGs) within a few milliseconds.

Accurate electricity demand forecasting is crucial for grid stability and resource optimization, yet predictive models degrade over time due to data drift caused by market fluctuations, policy changes, and infrastructure shifts. Undetected drift leads to forecasting errors, inefficiencies in energy distribution, and increased operational costs. This paper presents a deep learning-based framework for detecting data drift in electric load time series, leveraging statistical and machine learning-based approaches with a focus on the Pruned Exact Linear Time (PELT) algorithm for efficient change-point detection. Our method is scalable and highly efficient, making it suitable for real-time applications. Unlike traditional drift detection techniques, our approach dynamically adjusts to evolving patterns, mitigating both gradual and abrupt changes in consumption behavior. By leveraging both synthetically generated multivariate data and realworld univariate data for electricity load-related time series, the effectiveness of PELT is evaluated demonstrating its ability to capture structural shifts with high accuracy. The proposed framework achieved F1 score of 82% and 98% for generated temperature and humidity time series, respectively. Similarly, it achieved F1 score of 94% for real-world electricity demand data. We also assessed the quality of PELT’s detection by applying time series smoothing techniques that provided additional insights.

Urban intersections, critical hotspots for severe traffic crashes, demand advanced analytical approaches, particularly in developing nations where such methods remain underexplored. This study introduces a novel data mining framework integrating deep learning to analyze collision patterns at urban intersections in Qazvin, Iran. Utilizing 245 crash records (2021–2023), the research applied k-means clustering to categorize variables, regression trees for classification, and a deep convolutional neural network (DCNN) to evaluate data clusters. Three primary crash-influencing factors emerged: vehicle type, human factors, and lighting conditions. Training the DCNN on 2021–2022 data and testing it on 2023 data yielded 97% accuracy in predicting crash determinants. Findings highlighted unique crash characteristics across intersections, emphasizing context-specific risk factors. The model enables precise identification of variable importance, offering actionable insights for targeted safety interventions. This approach bridges methodological gaps in crash analysis for developing regions, demonstrating the efficacy of hybrid data mining and deep learning in enhancing intersection safety planning. By prioritizing key risk variables and enabling predictive analytics, the framework supports data-driven policymaking to mitigate crash severity and frequency in urban settings.

In recent years, photovoltaic energy has gained global attention due to the advantages of photovoltaic (PV) systems and the abundance of solar energy. Consequently, the installation capacity of PV systems has risen. Despite these benefits and the notable growth of PV systems, they face challenges such as high initial costs, low power conversion efficiency, reliance on environmental conditions, and vulnerability to faults. Fault detection in PV arrays is critical to minimize energy losses and maximize income for users while also improving electricity efficiency and system lifespan. However, because of the non-linear nature of PV systems, it is challenging for protection devices to detect faults, which can lead to safety risks and fire hazards in solar power plants. In this study, a convolutional neural network (CNN), a type of supervised model, was used for fault diagnosis. Yet, like other supervised models, it has several drawbacks: 1) Acquiring labeled PV data is costly and challenging. 2) Updating the trained model is difficult. 3) Visualizing the model is complex. To overcome these limitations, this study introduces a semi-supervised learning model that uses only a small amount of labeled data and normalizes it for better visualization.

Virtual Synchronous Generator can emulate the characteristics of a synchronous generator to provide inertia and damping for microgrid systems. In the case of employing a T-type three-level converter structure, the traditional model predictive control method has a large computational burden, and it is difficult to determine the weighting factor of the midpoint voltage. Additionally, although fixed-parameter VSG control can provide inertia and damping, it cannot guarantee frequency regulation performance. To address these issues, this paper proposes an optimal switching sequence model predictive control strategy for T-type three-level VSG with parameter synergistically adaptive control method. This method analyses the impact of different vectors output by the T-type three-level inverter on the midpoint potential and the charging and discharging status of the DC-side capacitor. By setting constraint conditions, the switching sequence of the inverter is optimized, eliminating the weighting factor used to control the midpoint voltage in traditional cost functions, resulting in smaller fluctuations in the DC-side voltage. Based on the active closed-loop transfer function of VSG, the optimal damping ratio is set, and a parameter synergistically adaptive formula is designed to allow coordinated adjustment of VSG parameters, effectively improving microgrid frequency stability during power fluctuations.

In recent years, due to the exponential growth in the applicability of biometric authentication systems, it has become essential to address the privacy and security concerns of user biometric information. The cancellable biometric template generation is one of the promising solutions in this situation which protects both the system and the user’s biometric data from unauthorized access. The complexity of such a system which sustains improved performance while maintaining the anonymity of users by introducing non-invertibility, unlinkability, and encapsulation of encrypted templates is a difficult task[1]. The random projection-based cancellable biometric template generation is one of the efficient techniques to secure the information of users. But still, this approach suffers from the adversary attack where an attacker can obtain the original template by performing repeated random projections on the stored template[2]. In this paper, we proposed a technique: MCC Encoded Random Triangle Hashing which protects the biometric template by encoding the projection matrix with the minutiae cylindrical codes obtained from the minutiae information. This technique addresses the problem where the projection matrix is leaked to an attacker which can lead to the reconstruction of the original biometric template. The performance of the proposed technique is evaluated using FMR, FAR, GAR, ROC Curve, and EER parameters on six FVC fingerprint databases FVC2000 DB1, FVC2000 DB2, FVC2002 DB1, FVC2002 DB3, FVC2004 DB1, and FVC2004 DB3 which is publicly available.

In the analysis of the operational reliability of integrated electrical energy systems, traditional numerical algorithms for solving natural gas dynamics require a substantial amount of computation, making it challenging to complete the analysis within operational time scales. This paper proposes a novel method: replacing traditional numerical methods with a CNN-LSTM neural network algorithm. The algorithm first utilizes a Convolutional Neural Network (CNN) for feature extraction, followed by a Long Short-Term Memory network (LSTM) for feature sequence prediction. Through a sequence-to-sequence learning process, the model learns the mapping relationships between adjacent time steps from the data, constructing a dynamic surrogate model for the gas network. This dynamic surrogate model is further integrated with the power system load flow model, combined with Monte Carlo simulation and multi-state models, to comprehensively analyze the operational reliability of integrated electrical energy systems. In the validation phase, the proposed model was applied to a distribution network-level electric-gas integrated energy system. The validation results demonstrate that the model not only accurately simulates the complex characteristics of gas network dynamics but also significantly reduces the computation time for operational reliability.