Power, Control, and Data Processing Systems

Abstract Because conventional computational approaches falter under heavily constrained and nonlinear conditions, properly setting protective distribution relays remains an arduous task. While the metaheuristic algorithms have been applied to this problem, existing approaches typically optimize only time and current settings while assuming fixed relay characteristics, limiting the achievable coordination performance. This paper addresses this gap by proposing a unified Gravitational Search Algorithm (GSA) framework that simultaneously optimizes time multiplier settings, current settings, and relay characteristic selection from eight standard curves. By leveraging a combination of MATLAB for optimization and DIgSILENT for fault current estimation, this study evaluates its approach. The effectiveness of the technique is demonstrated through testing on both 30-bus and 9-bus distribution systems, where its performance is benchmarked against Informative Differential Evolution and Particle Swarm Optimization. Results demonstrate that the GSA-based approach achieves superior convergence and lower objective function values, with statistical analysis confirming its robustness and effectiveness in reducing relay operating times compared to the existing methods.

Abstract This paper presents the design, implementation, and experimental evaluation of an autonomous line-following robot developed for efficient material transportation in dynamic industrial environments. Unlike conventional line-following robots that focus only on path tracking, the proposed system integrates automatic lighting control, onboard thermal management, and access security as core operational functions. These additional features enable the robot to maintain reliable performance under variable lighting and temperature conditions while preventing unauthorized use in shared workspaces. The navigation system is based on five TCRT5000 infrared sensors, which provide accurate line detection and stable maneuvering along complex trajectories, including sharp 90-degree turns. To enhance environmental adaptability, the robot automatically activates its lighting system in low-light conditions and employs a temperature-regulated cooling mechanism to protect onboard electronics and sensitive payloads. An access control layer is also incorporated to improve operational safety and reduce tampering risks. Experimental results demonstrate that the proposed robot achieves reliable path-following performance and robust subsystem response under different operating conditions. The combination of precise navigation, environmental adaptation, and security functions makes the system a practical and scalable solution for industrial automation and internal logistics in modern factory settings.

Abstract Accurate load forecasting is recognized as a critical requirement for the secure and economical operation of modern power systems. In this study, two advanced data-driven neural architectures are presented and compared for forecasting the load behavior of Iran’s power grid. In the first approach, a Time-Delay Cerebellar Model Articulation Controller (TD-CMAC) is employed to capture temporal dependencies in hourly load series, and its performance is systematically benchmarked against a conventional Multilayer Perceptron (MLP), Long Short-Term Memory (LSTM), and Extreme Gradient Boosting (XGBoost). In the second approach, a rough Gaussian Neural Network (RGNN), trained by the Levenberg–Marquardt algorithm, is developed, in which rough set concepts are combined with Gaussian activation functions to enhance robustness under uncertainty. This RGNN is evaluated on power load time-series data, so that its generalization capability across regression tasks can be assessed. Both neural frameworks are trained on real load data from the Iranian power network. Their performance is quantified using cost functions, regression plots, and accuracy metrics, demonstrating that while the proposed RGNN provides exceptional robustness against fluctuations, both architectures offer complementary advantages depending on the forecasting horizon. The empirical findings indicate that the TD-CMAC achieves competitive and often superior accuracy compared with the MLP, and XGBoost for short-term load forecasting, whereas the Levenberg–Marquardt-trained RGNN provides effective and robust modeling for prediction in the presence of uncertainty. Overall, it is demonstrated that the combination of time-delay cerebellar structures with rough Gaussian neural modeling constitutes a powerful data-driven toolkit for power load forecasting in complex power systems.

Abstract Predictive torque control (PTC) is known as a straightforward approach for induction motor drives for which the control parameters can be directly included in the criterion measure. Nevertheless, choosing the weight factors of the criterion measure is an important challenge. In conventional PTC methods, the criterion measure includes the torque and stator flux control objectives which have different weight factors for each control objective. Tuning weight factors empirically results in a complicated control process. To successfully cope with this issue, this paper suggests the VIKOR method for optimization of the criterion measure. In this research, a new PTC scheme has been proposed using the VIKOR optimization algorithm for the selection of inverter switching vectors connected to an open-end winding induction motor (OEWIM). The purpose of controlling the switching vectors is to reduce the flux and torque ripples. The results are then compared with the conventional PTC and ranking methods. It is demonstrated that, with optimization by the VIKOR method, the flux and torque ripples have been remarkably reduced in comparison with the existing approaches. The proposed VIKOR-based PTC reduces the flux ripple by 78.5% (from 1.4 Wb to 0.3 Wb) and reduces the torque ripple by 58.3% (from 0.012 N·m to 0.005 N·m) compared to the conventional PTC method. Finally, a step-by-step implementation algorithm is presented and simulated using MATLAB which demonstrates favorable results.

Abstract The automated attribution of historical paintings to their school of origin is a long-standing challenge at the intersection of computer vision and art history. This challenge is especially acute for non-Western art traditions, such as Persian miniature painting, for which digitized and labelled image corpora remain comparatively small. This paper addresses the classification of Persian miniature paintings into five historical schools including, Herat, Qajar, Shiraz-e Avval, Tabriz-e Avval, and Tabriz-e Dovvom, using a Vision Transformer (ViT). Transfer learning is used to adapt a ViT-B/16 model, pre-trained on ImageNet-21k, to this task. Because the available dataset contains only 890 images, five complementary augmentation operations i.e. random-resized cropping, horizontal flipping, rotation, color jitter, and random erasing, are applied during training, together with Mixup as an additional sample-mixing regularizer. A progressive layer-freezing strategy keeps the first six Transformer encoder blocks fixed and fine-tunes only the last six blocks. To stabilize training on the small fine-tuning set, this strategy utilizes a classification head combined with label smoothing and a low, cosine-annealing learning rate. On a held-out test set of 178 images, the fine-tuned model reaches an overall accuracy of 92.13% and a weighted F1-score of 0.92, with three of the five schools classified at an F1-score above 0.94. These results indicate that a Vision Transformer, can reach competitive accuracy on a small, specialized cultural-heritage dataset when transfer learning and augmentation are used simultaneously.

Abstract This paper proposes a deep operator learning framework for fast and accurate prediction of both primal and dual variables in alternating-current optimal power flow (AC-OPF). Unlike conventional neural-network surrogates that learn a fixed input--output mapping and often ignore dual information, the proposed method is designed to approximate the underlying operator that maps varying system operating conditions to optimal OPF solutions. The framework is built on a primal--dual DeepONet architecture, where branch and trunk networks jointly encode system conditions and query locations to predict generator dispatch, voltage states, and associated Lagrange multipliers.

To better capture optimization structure, a differentiable dual completion layer is incorporated to improve dual feasibility and complementary slackness, and a KKT-informed loss function is used to guide training toward physically consistent and optimization-aware solutions. The proposed model is evaluated on standard IEEE benchmark systems, where it achieves high prediction accuracy for both primal and dual variables, millisecond-level inference time, and significantly reduced KKT residuals compared with conventional learning-based OPF approximations. These results highlight the promise of deep operator learning as an efficient surrogate for real-time AC-OPF analysis and power-system operation.