https://eej.aut.ac.ir/ AUT Journal of Electrical Engineering en daily 1 Wed, 01 Jul 2026 00:00:00 +0330 Wed, 01 Jul 2026 00:00:00 +0330 https://eej.aut.ac.ir/article_5997.html In general, symmetrical and asymmetrical capacitor-clamped boost converters and direct current capacitor voltage, unbalancing specially with lower output current Total Harmonic Distortion are frequent problems for inverters. In order to improve voltage quality, a boost converter with a flying capacitor and grid tie inverter is suggested in this study. It manages the direct current link voltage asymmetrically. Additionally, the boost converter with flying capacitor grid tie inverter system's output voltage dynamic responses are enhanced and simulated using MATLAB Simulink which in turn benchmarked using a scaled-down hardware module. Proportional Integral and Model Predictive Controller control strategies are suggested and implemented in the built hardware. The suggested system's voltage, current, and dynamic performance are examined. The results show that a 360 Watt output power can be delivered by the suggested combination of the described converter system. Additionally, grid-connected power converters and flying capacitor boost converters have lower current harmonics and better voltage regulation direct/alternating current converters, demonstrating the developed system's great suitability for power usage in home photovoltaic systems. https://eej.aut.ac.ir/article_6018.html Currently, the combination of Renewable Energy Sources (RES), particularly Photovoltaic (PV) systems, into power networks has grown in importance for sustainable energy generation. Therefore, this research develops the control approach for a high gain Triple Winding Max Gain Boost (TWMGB) converter incorporated with a Maximum Power Point Tracking (MPPT) controller for PV systems. The developed converter exploits a triple winding inductor structure to attain an improved voltage gain, making it appropriate for low voltage PV system needs effective step-up capability. An innovative MPPT control approach based on Walrus Optimization Algorithm (WOA) tuned Radial Basis Function Fuzzy Inference System (RBFFIS) is utilized to extract the upmost power from the PV array in dynamic ecological conditons. It assures fast convergence to the globalMPP and enahnces tracking accuracy even in partial shading scenarios. Moreover, the coordinated interaction among the TWMGB converter and adaptive control approach assures better performance interms of diminshed voltage stress and ripple. The performance of a system is applied via MATLAB/Simulink tool, demonstrating its adaptability and robustness with converter efficacy of 97.61%. The developed system offers a consistent and scalable solution for advanced PV based power systems, contributing to sustainable energy conversion and utilization. https://eej.aut.ac.ir/article_6029.html This work presents an intelligent Photovoltaic (PV) grid integration system employing a Northern Goshawk Optimization Algorithm (NGOA)-based Radial Basis Function Neural Network (RBFNN) for Maximum Power Point Tracking (MPPT) and Soft Clamp X-Gain Boost (SC-XGB) converter for enhanced voltage regulation. The proposed system aims to optimize energy harvesting from PV source and ensure stable power delivery to a three-phase grid. The RBFNN is trained offline using comprehensive PV datasets and directly predicts the optimal duty cycle from measurable PV inputs during real-time operation, eliminating the need for auxiliary MPP pre-estimation algorithms, while NGOA enhances RBFNN’s learning capability by fine-tuning its weights and biases for rapid and accurate MPPT performance even under varying irradiance and temperature conditions. PV output is connected to SC-XGB, which efficiently raises the Direct Current (DC) voltage and is thus controlled by Pulse Width Modulation (PWM) signals which are generated according to MPPT output. Regulated DC output is then transformed into a three-phase Alternating Current (AC) by a Voltage Source Inverter (VSI), the output of which is taken through an LC filter to reduce harmonics before the power is fed into grid. Simulation is done in MATLAB showing the capacity of NGOA-RBFNN to track Maximum Power Point (MPP) at very high speed and accuracy. The system achieves superior voltage regulation of 95.24% efficiency, reduced Total Harmonic Distortion (THD) and enhanced dynamic performance over traditional MPPT control methods. https://eej.aut.ac.ir/article_6019.html At present power system faces certain Power Quality (PQ) issues, due to large amount of power usage, fluctuations and other uncertainties in non-linear loads. Thus, Distribution Static Compensator (DSTATCOM) is deployed for mitigating PQ problems. A three-phase AC source system, supplying a non-linear load using parallel-Voltage Source Inverter (VSI) based DSTATCOM at Point of Common Coupling (PCC) is deployed to DC-Link which functions based on current sharing principle. A novel Re-lift Single Ended Primary Inductor Converter (SEPIC) converter is integrated with Photovoltaic (PV) to boost PV power generation, assuring consistent and sustainable power supply to DC-Link capacitor of DSTATCOM is correctly charged. To further enhance the system performance, D-Q theory/Neural Network-based Synchronous Reference Frame (SRF) theory is utilized for generating reference current for DSTATCOM. These control topologies enable accurate compensation of reactive power and harmonic currents in real-time, assuring improved grid voltage stability and rectifying distortions. Proposed system is executed using MATLAB simulation and acquired outcomes validate improved system functioning with better PQ mitigation at PCC under varying load conditions. Thus, demonstrating the impacts of integrating renewable energy with advanced control approaches. https://eej.aut.ac.ir/article_6008.html This paper addresses voltage stability challenges in distribution networks through a coordinated approach using Distributed Generation (DG) and Static Synchronous Compensator (STATCOM) placement. A novel hybrid Grey Wolf Optimization-Particle Swarm Optimization (GWO-PSO) algorithm is proposed to optimize the placement and sizing of these components with the objective of enhancing static voltage stability. The Fast Voltage Stability Index (FVSI) is employed as the primary metric for assessing voltage stability, where lower values (approaching zero) indicate improved stability. The proposed hybrid algorithm leverages GWO's exploration capabilities and PSO's exploitation strengths to overcome the limitations of individual algorithms. The methodology is validated on a 35-bus distribution system with a total load demand of 1.89 MW and 1.3455 MVAr. Results show that the hybrid GWO-PSO achieves an average FVSI reduction of 16.98%, significantly outperforming both standalone GWO (12.45%) and PSO (14.32%) implementations. The voltage profile across all buses is substantially improved, with the hybrid approach maintaining voltages closer to nominal values of 1.0 p.u. compared to the base case where many buses operate under low voltage conditions. The hybrid algorithm demonstrates faster convergence, reaching optimal solutions within 100 iterations compared to individual GWO and PSO implementations. The coordinated placement strategy determined optimal DG and STATCOM sizes and locations, effectively addressing voltage stability concerns in distribution systems experiencing rapid load growth with insufficient reactive power support. https://eej.aut.ac.ir/article_6022.html In most areas and power systems, Photovoltaic (PV) energy is rapidly becoming a significant component of the energy balance because of its rapid annual growth rate. Therefore, this research presents the PV-fed improved SEPIC-Zeta converter with a cascaded fuzzy algorithm based Maximum Power Point Tracking (MPPT) for efficient DC power transfer. At the beginning, the improved SEPIC-Zeta (ISZ) converter is exploited to enhance the PV system’s voltage. Then, the Cascaded Fuzzy MPPT algorithm is introduced for tracking the upmost power from PV system. Also, the high frequency inverter transmutes the DC to AC power and isolation is provided by the isolation transformer for ensuring safety and mitigating harmonic distortion on the source and load side. Additionally, the interleaved synchronous rectifier is exploited for converting the AC into a DC supply. The implemented research is validated in MATLAB tool, which demonstrates that the proposed work has a converter efficacy of 95.12 %, which handle fluctuations and disturbances more effectively, enhancing the reliability of overall system. https://eej.aut.ac.ir/article_6009.html WECS are dynamic and intricate, considered by uncertainties and external disturbances. This research introduces an innovative pitch control strategy designed to improve energy stabilization and extraction predominantly for WT affected by unmodeled system dynamics to ensure stable operation under high wind speeds. For optimizing power capture, the ANN controller adapts dynamically to changing wind conditions by regulating the turbine blade’s pitch angle. For grid integration, a PWM rectifier transforms the variable-frequency AC power from the turbine into DC power. The simulations are conducted in MATLAB/Simulink tool to evaluate the control framework. It reveals that the superior performance of the ANN controller in reducing mechanical loads on turbines and platforms and maximizing power generation. Thus, it achieving reduced power and speed fluctuations, minimal overshoot and enhances the dynamic behaviour of WT. https://eej.aut.ac.ir/article_6010.html Accurate estimation of battery parameters, particularly State of Charge (SOC) and State of Health (SOH), is critical for the operational reliability and safety of electric vehicles (EVs). These parameters influence driving range, charging strategy, and long-term battery lifespan. Traditional methods such as Coulomb counting, equivalent circuit models, and Kalman filters have been standard for battery state estimation but struggle with noisy data, variable loads, and nonlinear battery ageing. Recently, deep learning has shown promise in addressing these challenges by offering more robust and adaptive performance.A recent review proposes a 4C framework—Correctness, Compute, Calibration, and Compliance—to evaluate deep learning models for next-generation Battery Management Systems (BMS). This scheme prioritises practical deployment aspects alongside accuracy. The review covers over 60 studies from 2019 to 2024, assessing model architectures, input features, training methods, and deployment readiness. It highlights advances such as physics-informed and uncertainty-aware models and offers a comparative evaluation of accuracy and computational efficiency on public datasets.Deep learning methods consistently outperform traditional approaches, achieving SOC errors below 2% and SOH deviations within ±3%. Transformer-based and hybrid models improve accuracy by 10–20% compared to simpler recurrent models. Lightweight architectures like GRUs offer fast inference (less than 20 milliseconds), suitable for in-vehicle real-time applications.Despite promising results, challenges remain around data generalizability, explainability, and real-time deployment. The 4C framework offers a roadmap for bridging laboratory advances with reliable, production-ready BMS technologies. https://eej.aut.ac.ir/article_5995.html Unexpected failures in essential industrial systems can cause operational disruptions and financial losses. To mitigate unplanned downtime and maintain safe, efficient functioning of critical assets, predictive maintenance strategies are essential. However, with the rapid increase in sensor-equipped machinery, the overwhelming volume of generated data has outpaced the capabilities of traditional machine learning models to provide accurate, real-time diagnostics. This research introduces a model-free deep reinforcement learning (DRL) approach tailored for predictive maintenance within sensor-integrated equipment networks. Each machine is equipped with a sensor module that captures real-time data and detects anomalies. Unlike conventional opaque regression-based methods, the proposed framework autonomously determines optimal maintenance policies and delivers actionable insights for each individual device. Experimental evaluations indicate the potential of this adaptive learning method to extend across diverse maintenance scenarios. https://eej.aut.ac.ir/article_6043.html Three-dimensional Networks-on-Chip (3D-NoC) application mapping is a nondeterministic polynomial-time hard problem by nature. Under tight design restrictions, the efficient allocation of Intellectual Property (IP) cores to processing components must balance silicon area, reduce power consumption, and reduce end-to-end latency. To achieve improved mapping quality for 3D-NoC designs, this paper presents a Hybrid Lemur–Dwarf Mongoose Optimisation (LDMO) approach that addresses the cooperative exploitation abilities of Dwarf Mongoose Optimisation and the exploratory behaviour of Lemur Optimisation. To prevent premature stagnation, the Lemur Optimisation Algorithm mimics lemur cliff-leaping and tree-navigating behaviour during first stage. This produces a diversity of initial mapping candidates with high population variance. To optimise the mapping to global optima with high convergence speed, the Dwarf Mongoose Optimisation Algorithm employs adaptive leadership, sentinel–scout coordination, and foraging-based neighbourhood search during the second stage. The average communication delay (hop-dependent propagation, serialisation, and router latencies), area overhead (switch, interconnect, and core dimensions), and total power dissipation (router and interconnect power) are all minimised at the same time through a multi-objective fitness function. By adaptive switching between the two phases, the hybrid approach dynamically trades off between exploration and exploitation to ensure robustness across a wide range of communication demands and traffic patterns. Simulation outcomes demonstrate that the proposed LDMO framework consistently delivers reduced computation overhead, marked improvements in latency, and substantial energy efficiency. Furthermore, as the number of cores and communication links scale upward, the hybrid optimisation strategy maintains high-quality mapping solutions, underscoring its robust scalability across diverse NoC configurations. https://eej.aut.ac.ir/article_6030.html This work presents a compact, hexagon-shaped wideband antenna designed for wearable applications in the Industrial, Scientific, and Medical band. The antenna is fabricated on a flexible felt substrate with a dielectric constant of 1.22 and a loss tangent of 0.016, achieving a compact footprint of 35 × 30 × 2 mm³. A Coplanar Waveguide feed integrated with a Stepped Impedance Resonator is employed to enhance impedance matching and bandwidth performance. The felt substrate exhibits excellent mechanical durability and electromagnetic stability under repeated bending and typical wear conditions, ensuring reliable long-term operation. The proposed antenna achieves a peak gain of 7.31 dB at 5.82 GHz and maintains a Specific Absorption Rate of 1.08 W/kg for 1 g of tissue, which is well within regulatory safety limits. Beyond communication, the antenna demonstrates moisture-sensing capability, exhibiting a consistent downward shift in resonant frequency with increasing substrate humidity. While stable under bending, excessive moisture leads to detuning and impedance mismatch. The strong correlation between simulated and measured results validates the proposed design as a robust and multifunctional solution for wearable biomedical and ultra-wideband sensing applications. https://eej.aut.ac.ir/article_6033.html The optimal design of a Switched Reluctance Motor (SRM) requires an accurate model. Analytical models presented for SRM do not meet required accuracy. Approximate models also have varying degrees of accuracy in predicting SRM characteristics. Hence, in this paper, two Neural Network (NN) models, Radial Basis Function (RBF) and Multilayer Perceptron (MLP), have been proposed to predict torque ripple and average torque, respectively. To train and test the models, 100 samples were extracted, 90% for training and 10% for testing. Furthermore, the Finite Element Method (FEM) has been used to solve the samples. The influencing parameters of the proposed NN models (number of hidden layers, number of neurons in hidden layers, bias, etc.) also have been determined to achieve the desired accuracy and minimal complexity. To evaluate the performance of the models, two criteria, Root Mean Square Error (RMSE) and Mean Relative Error (MRE), have been used. Both criteria indicate that the MLP model is successful in predicting torque ripple, while the RBF model excels in predicting average torque. https://eej.aut.ac.ir/article_6063.html This paper presents a novel framework that integrates secure, distributed computation with advanced control to enable confidential and collision-resilient formation flight for multiple quadrotors. Conventional multi-agent systems are vulnerable to cyber threats such as eavesdropping and data manipulation, which can destabilize formations and compromise missions. In this study, to address this issue, Model Predictive Control is combined with a multi‑party computation protocol based on secret sharing. In this approach, each quadrotor converts its state data into secret shares and distributes them among the agents, and the required control computations are performed collaboratively on the encrypted data without revealing any sensitive information. This decentralized approach eliminates the central controller as a single point of failure and attack. Simulation results for six quadrotors flying in a circular formation among obstacles show that the system maintains formation accuracy and collision avoidance while remaining robust against eavesdropping and data manipulation. Quantitative evaluation shows that the formation tracking error remains below 0.08 m, all inter-agent and obstacle collisions are successfully avoided (100% success rate), and the secret-sharing protocol reconstructs the optimal control signals with a normalized mean squared error of less than 10⁻⁶, ensuring cryptographic integrity without compromising formation performance. The integration fundamentally enhances operational security, preserves agent privacy, and improves system fault tolerance and scalability for real-world adversarial environments. https://eej.aut.ac.ir/article_6064.html Currently, Photovoltaic (PV) solar power generation is integrated into distributed generation (DG) systems is becoming popular. Therefore, this research proposes a unique Coupled inductor based multiport (CIM) converter that combines battery storage and solar PV in a grid-connected system. The PV system produces inferior voltage owing to its intermittent and sporadic nature due to the climatic conditions. A CIM converter is utilized to convert the PV output voltage to the highest level. In order to combat the intermittency and instability associated with PV, a Proportional Integral (PI) controller is added to the grid to ensure a stable and uninterrupted supply of power. Also, the battery is exploited to store surplus energy from PV systems, permitting to use this stowed energy when the panels aren’t producing electricity. A Three phase Voltage Source Inverter (3ϕ VSI) is used to transform steady DC to AC voltage and a LC filter is employed to lessen harmonics. Additionally, the grid voltage synchronization is carried out by PI controller on load side. The developed system is also tested with the simulation outcomes from MATLAB/Simulink and comparison is made with conventional approaches. The obtained outcomes show that it attains the efficiency of 97.46%, ensures the optimized power is delivered to the grid. https://eej.aut.ac.ir/article_6065.html This study presents a diagnostic and regrouping approach for used 18650 lithium-ion cells using the DBSCAN clustering algorithm. A total of 154 cells were recovered from discarded laptop batteries. After electrical testing (constant-current discharge measuring capacity, voltage, and internal resistance), 120 cells (78%) were identified as healthy and suitable for reuse. The DBSCAN algorithm (eps=0.3, min_samples=10) was applied to cluster these 120 cells based on their electrical characteristics. The algorithm successfully formed three homogeneous clusters of 40 cells each. These clusters were assembled in a 3S40P configuration (three parallel packs of 40 cells connected in series) to form a second-life battery of approximately 81 Ah capacity at 10.9 V. Comparative analysis shows that DBSCAN outperforms K-means and hierarchical clustering for this application, achieving a silhouette coefficient of 0.62 versus 0.48 for K-means. The proposed method achieves a 78% cell recovery rate, demonstrating its effectiveness for battery recycling and second-life applications. https://eej.aut.ac.ir/article_6069.html Dynamic cloud and edge-cloud platforms require task schedulers that can respond to stochastic workloads while minimizing energy use and preserving service-level agreement reliability. This study proposes an attention-driven deep reinforcement learning framework for energy-efficient and service-level agreement-aware cloud task scheduling. The framework combines lightweight convolutional neural network and long short-term memory-based spatial-temporal feature extraction with a multi-head self-attention actor-critic decision module. The convolutional neural network and long short-term memory components capture local virtual machine workload patterns, whereas self-attention models global virtual-machine-to-virtual-machine dependencies for parallel and context-aware scheduling. The scheduling problem is formulated as a Markov decision process using a 242-dimensional virtual-machine-level state representation, probabilistic virtual-machine-to-host assignment actions, and a multi-objective reward function covering makespan, energy consumption, operational resource cost, and service-level agreement penalties. Experiments were conducted in a heterogeneous CloudSim environment with 100 hosts and 100 virtual machines. The proposed framework achieved a normalized makespan of approximately 0.85, a 14.4% reduction in total energy consumption, consistently low service-level agreement violation behavior, and controlled migration activity. Logged analysis further showed a response time of 10.0000 milliseconds per completed task or virtual machine event, supporting interval-based real-time feasibility. Cost is treated as an operational reward component, not as a standalone billing analysis. https://eej.aut.ac.ir/article_6091.html Given that the DC-DC buck converter (DDBC) is an example of a nonlinear system, and also considering the disturbance and changes related to its parameters, the output voltage control in this DDBC will face several challenges. In this paper, a PD-PI cascaded controller is designed to control the voltage in the DC-DC buck converter. In this cascaded controller, the PD controller, as the outer loop controller, is responsible for responding quickly to voltage changes and damping them, and the PI controller is responsible for eliminating the error caused by voltage changes. Considering that the accurate and optimal adjustment of the parameters related to the cascaded controller is crucial to its performance, the Grey Wolf Optimizer algorithm has been employed to optimize these parameters. This algorithm offers several advantages, including intelligent search, parameter fine-tuning, and stable and rapid convergence. To assess the robustness of the proposed PD-PI(Grey Wolf Optimizer) approach against disturbances and parameter variations, multiple scenarios were developed and its performance compared with that of control strategies like PD-PI (Electric Eel Foraging Optimizer) and PD-PI (Flood algorithm). According to the results, the proposed method outperforms other control methods in terms of various criteria, including Settling time, Rise time, Peak time, Overshoot, and Steady state error. https://eej.aut.ac.ir/article_6105.html This paper presents a novel control strategy that combines RST-based speed control with Direct Thrust Force Control to improve the operational performance of linear induction motors. This approach specifically targets the optimization of thrust force, magnetic flux, and current control. While traditional PID controllers are commonly used in direct torque control applications for rotary motors, their effectiveness in linear induction motors is limited by sensitivity to parameters and nonlinear effects caused by end-effect phenomena. These limitations frequently result in excessive overshoot, slower rise and settling times, as well as reduced control accuracy and system stability. In response, this study proposes the RST- Direct Thrust Force Control method, which is designed to achieve robust speed regulation and precise control over thrust force and flux under various operating conditions. The performance and accuracy of the proposed approach are evaluated and validated through simulations conducted in the MATLAB/Simulink environment. Moreover, comparative simulation analyses demonstrate that the RST- Direct Thrust Force Control technique significantly reduces overshoot, shortens rise and settling times, and confirming its superiority in maintaining control precision, including excellent reference tracking and disturbance rejection. The findings highlight the RST regulator combined with the Direct Thrust Force Control technique as a robust solution for overcoming the inherent limitations of conventional direct torque control in linear induction motor applications. This integrated control presents a promising approach to optimizing linear induction motor performance and aligns with industry demands for precision and reliability in dynamic environments. https://eej.aut.ac.ir/article_6106.html Brain tumor segmentation from Magnetic Resonance Imaging is a critical step for diagnosis and treatment planning, yet manual delineation is time-consuming and subjective. While deep learning models like Convolutional Neural Networks and Transformers have advanced automated segmentation, they face inherent limitations. CNNs struggle with long-range dependencies, while pure Transformers are often data-inefficient. To address these challenges, we propose DASA-Net, a novel hybrid Dual-Attention Swin-Transformer Architecture for precise brain tumor segmentation. Our model integrates a pre-trained Swin Transformer as the encoder to capture rich hierarchical and global contextual features. We introduce an Enhanced Spatial Attention module with residual connections to refine spatial details in the encoder's feature maps, and incorporate Squeeze-and-Excitation blocks in the decoder for adaptive channel-wise feature recalibration. Additionally, a Dynamic Hybrid Loss function is employed to balance Binary Cross-Entropy, Dice, and Focal losses, effectively addressing class imbalance and boundary ambiguity. Extensive experiments on the BRISC and Figshare brain tumor datasets demonstrate that DASA-Net achieves state-of-the-art performance, with a Dice score of 90.3% and IoU of 82.3% on BRISC, significantly outperforming existing methods. https://eej.aut.ac.ir/article_6117.html This brief discusses the stabilization of a flexible satellite's attitude while considering actuator dynamics. It employs a constrained Lyapunov Nonlinear Model Predictive Control (LNMPC) method combined with the Feedback Linearization Technique (FBL). FBL is chosen for its ability to cancel the system's nonlinearity, making the control law easier to implement and more effective. more over the LNMPC is not able to nullify the low frequencies so it is necessary to combining it with a nonlinear controller. This composite controller aims for high accuracy in tracking the reference trajectory and, due to the LNMPC, provides tremendous robustness against unknown disturbances. The hybrid controller is designed to steer the satellite's attitude and bring the reaction wheel's angular momentum to zero to save energy. This paper addresses various challenges faced by satellites during their missions, including uncertainties, external torque disturbances, actuator faults, and vibrations of appendages. The performance of the proposed controller is analyzed while considering all these challenges. The active set method is used as the optimization algorithm, more over a nonlinear disturbance observer named Cascade Extended State Observer ( Cascade ESO) is introduced to overcome the external disturbance and improve the performance of the controller and closed-loop stability is addressed using a candidate Lyapunov function. https://eej.aut.ac.ir/article_6118.html The analysis of degraded ancient documents remains challenging because multiple degradation patterns may appear within a single document image and reduce document quality. In this paper, we propose a combination of convolutional neural networks and extreme learning machine for patch-level degradation classification in ancient document images. The dataset was constructed from 100 source document images and cropped into 700 labeled patches. Four pre-trained architectures, namely VGG19, ResNet101, MobileNetV2, and EfficientNetB0, were used as feature extractors. The number of hidden nodes in the ELM classifier was tuned, and two activation functions, namely radial basis function and sigmoid, were evaluated. The experimental results show that EfficientNetB0 features combined with the extreme learning machine classifier achieved the best performance, with a testing accuracy of 98.21% ± 2.08%. This combination also produced faster inference time compared with the other evaluated models. These results indicate that the EfficientNetB0-extreme learning machine combination is effective for patch-level degradation classification in ancient document images. https://eej.aut.ac.ir/article_6120.html Power quality issues have emerged as one of the most significant problems facing a power system, due to its diverse nature and frequent occurrence. This research introduces a Photovoltaic (PV) incorporated Unified Active Power Filter (PV-UAPF) system and its control approach for generation of decarbonized clean energy and improved Power Quality (PQ). The UAPF, combining shunt and series converter, which is utilized for compensating the current quality and voltage quality issues. Moreover, the PV system is incorporated with Switched Inductor (SI) based Boost-Luo converter at DC link, which is connected between series and shunt converters. For control the operation of converter, PI controller is employed and its parameters are refined by Modified Red Kite Optimization (MRKO) algorithm. For ensuring efficient voltage regulation and to manage the functioning of UAPF, an Artificial Neural Network (ANN) controller is incorporated on load side. To further demonstrate developed work's efficacy, it is implemented in MATLAB/Simulink and comparison is made with existing methods. The outcomes reveals that the developed system has a THD value lower than 5%, ensures the quality of power is enhanced. https://eej.aut.ac.ir/article_6121.html Short-term electricity consumption forecasting plays a crucial role in the optimal operation of energy systems and smart grids. Despite the widespread success of machine learning models in improving forecasting accuracy, their lack of interpretability often hinders trust and application in energy decision-making. This paper presents an interpretable framework for hourly electricity consumption prediction using lagged load and temperature features. To this end, the performance of linear and non-linear models is systematically compared on a one-year hourly electricity consumption dataset, including ambient temperature and workday information for a commercial building. To enhance model transparency, Explainable Artificial Intelligence methods are employed to analyze feature importance at both global and local levels. Results indicate that short-term and daily lags of electricity consumption play a dominant role in consumption dynamics, while the effect of temperature is non-linear and often appears with a time delay. Tree-based models significantly outperform the linear model and demonstrate a higher ability to identify complex interactions and lagged effects of temperature. SHapley Additive exPlanations analysis reveals consistent patterns in feature importance, and its waterfall plots and Local Interpretable Model Agnostic Explanations provide complementary local insights under different consumption conditions. The proposed framework, while improving forecasting accuracy, provides a transparent and meaningful interpretation of electricity consumption behavior and can be used as a reliable tool in energy planning and demand-side management. https://eej.aut.ac.ir/article_6122.html This paper presents a robust backstepping control strategy for Quanser’s 2-Degree-of-Freedom (2-DOF) helicopter, addressing challenges posed by external disturbances and actuator saturation constraints. The proposed approach leverages Flower Pollination Algorithm (FPA) to fine-tune the control parameters, enhancing system stability and performance. Initially, a mathematical model of the 2-DOF helicopter is derived, incorporating non-linear dynamics and actuator limits. The backstepping control framework is then developed, integrating disturbance rejection mechanisms to mitigate external perturbations. FPA is employed to optimize the controller gains, ensuring optimal performance under various operational scenarios. Extensive simulation results demonstrate the effectiveness of the proposed method, displaying significant improvements in tracking accuracy and robustness against disturbances and actuator saturation compared to conventional control techniques. This study shows the potential of combining the proposed metaheuristic optimization algorithm with robust control strategy for complex aerospace systems like helicopter. https://eej.aut.ac.ir/article_6123.html The growing demand for energy and the need to reduce emissions have made hybrid renewable energy systems (HRES) a strategic priority. Among these systems, the integration of solar, wind, and fuel cell resources in a complementary structure offers significant potential for reliable and low-emission energy supply. However, optimal site selection for deploying such systems – particularly in countries with diverse climatic and contextual conditions like Iran – presents a multi-criteria challenge accompanied by inherent uncertainty. This study proposes an integrated decision-making framework for prioritizing solar–wind–fuel cell hybrid systems across Iran’s 31 provincial capitals. The technical, economic, and environmental performance of the hybrid system was simulated using HOMER, generating indicators such as Cost of Energy (COE), Net Present Cost (NPC), CO2 emissions, renewable energy share, and capacity factors. These indicators, along with contextual criteria like population, natural disaster risk, and land price, were integrated into the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to rank the provinces. To assess robustness, an uncertainty analysis based on Monte Carlo simulation (10,000 iterations) was conducted. Results indicate that Tehran Province exhibits the highest closeness coefficient (~0.69) with a greater than 90% probability of being ranked among the top five. Conversely, regions such as Kerman and Sistan and Baluchistan consistently obtained lower ranks across most scenarios. This study demonstrates that integrating TOPSIS with Monte Carlo simulation provides not only quantitative rankings but also a risk-informed decision-making tool, effectively supporting policymakers in the siting of pilot renewable energy projects at the national scale. https://eej.aut.ac.ir/article_6126.html Prediction of solar energy generation is crucial in providing stable electrical supply and integrating renewable energy onto the electrical grid. To address challenges associated with complexity of the data, efficient modeling for the data and achieving interpretability of the predictions is essential. This paper presents a hybrid Deep Learning (DL) model for forecasting solar energy generation. The method initiates with a complete data preparation process for solar generation data by using the IQR method for robust outlier identification, followed by a thorough analysis of the distribution and variability in the data to determine trends in the data. The forecasting model for solar generation is developed using a new architecture that combines a Neuro Sparse Attention Integrated Liquid State Machine with Cross Latent Transformer Networks (NS-LSM-CLTNet), Hawkfish Optimization (HFO) algorithm is implemented to find the optimal configuration of the deep neural networks for ensuring the best possible performance from the trained models. The proposed NS-LSM-CLTNet model's forecast accuracies are validated in Python and achieves highest R² of 0.99 and lowest error rates such as MSE of 1306.94, RMSE of 36.151 and MAE of 21.602, which indicate that the proposed model performs significantly better than the existing approaches. Additionally, SHapley additive explanations (SHAP) framework is used to enable an explainable model. As a result, this provides certainty and transparency to grid operators and energy stakeholders about the forecasts produced by the system, thereby eliminating the black-box quality associated with many forecast solutions. https://eej.aut.ac.ir/article_6127.html Environmental benefits of Electric Vehicles (EVs) are associated with reduction of direct air pollutants and reduction of greenhouse gases. In this paper, a novel solar-powered EV Charging Station (EVCS) is offered with the integration of a grid-based energy management system and a high-efficiency Double-Stage Coupled Inductor Boost Converter (DSCIB). A new Arctic Puffin Optimized (APO) Proportional-Integral (PI) controller is employed for preserving a constant DC-link voltage and optimal system performance. This approach optimizes the energy transfer efficiency and enhances a stability under dynamic operating conditions. DC bus is linked to electrical grid through LC filter and three phase Voltage Source Inverter (VSI), which allows bidirectional power flow and reactive power support. PI-based control loop is employed for grid synchronization, to properly balance the active and reactive power. The charging and discharging processes of EV battery are controlled via a bidirectional DC-DC (BDC) converter, which is structured with the PI control strategy to precisely control the voltage and current. This configuration offers the capability of battery-to-grid without any interruption and fast charging of EV batteries during excess PV generation. The MATLAB simulation results shows proposed architecture offers a promising option for sustainable EV integration in grid applications because of the coordinated control framework. From simulation converter attain efficiency of 98.4% and decrease THD of 1.02% ensure effective energy flow, voltage stability, and grid support. https://eej.aut.ac.ir/article_6128.html Wireless power transfer (WPT) enables the transmission of electrical energy without physical connections, commonly used in charging devices and electric vehicles. Conventional methods often face limitations including narrow range, inadequacies in energy transfer, and high costs associated with infrastructure. In order to overcome the issues, this research proposes a Bridgeless Power Factor Correction Active Boost Rectifier (PFC-ABR), aimed at enhancing the efficiency of WPT for charging applications. The proposed architecture employs an AC source connected to a source inductance, which transfer power into a Bridgeless PFC-ABR that efficiently converts AC to DC power. The rectified output is processed by a single-phase high-frequency inverter, generating high-frequency AC power that is transmitted wirelessly through an isolation transformer. On the secondary side of the transformer an interleaved synchronous rectifier is provided to obtain a stable DC output, supplying power to the load through a capacitor bank. The experimental results are implemented by using MATLAB SIMULINK software which displays the efficiency as 93% with reduced THD 0.49% along with unity power factor. The proposed framework, demonstrates significant potential towards improving the performance of wireless charging systems, making it highly suitable for future high-efficient and high-power wireless energy applications. https://eej.aut.ac.ir/article_6129.html The growing proportion of Renewable Energy Sources (RESs) poses significant challenges for modern grids, particularly in terms of power flow stability, voltage management, and power quality. Traditional UPFCs rely on grid-connected DC link limiting their ability to operate sustainably and adapt flexibly to changing system conditions. To overcome such issues this research proposes a Photovoltaic (PV)-powered UPFC system incorporated with a Coupled Inductor (CI) Active Boost Step Up Converter to provide a stable high-gain supply to DC link. An advanced control strategy is introduced to extract maximum power using Artificial Neural Network (ANN) based Sharp belly Fish Optimization Algorithm (SFO) under varying environmental conditions. The series converter is connected through a transformer operates as a dynamic voltage compensator and shunt converter combined with an LC filter performs DC voltage regulation. MATLAB/Simulink validation shows better performance in terms of reduced Total Harmonic Distortion (THD), faster transient response, and enhanced steady-state voltage control as compared to traditional UPFC techniques. The obtained efficiency is around 94% and tracking efficiency is 99.9% with reduced THD 0.75%. https://eej.aut.ac.ir/article_6130.html Industrial manufacturing lines are well suited to integrated multi‑energy planning; however, many existing energy hub (EH) studies rely on synthetic load profiles or simplified thermal representations that limit their practical applicability at the factory scale. This paper develops a measurement‑informed, factory‑scale EH optimization model for a fabric‑reinforced concrete production line. Grid electricity and natural gas are modeled as primary energy carriers, while photovoltaic (PV) generation, diesel backup generation, and furnace waste‑heat recovery (WHR) are incorporated as complementary technologies within a transparent linear programming framework. All model parameters are derived from on‑site electrical and thermal audit data, allowing the analysis to reflect the actual operating envelope of the production line. Seven scenarios are investigated to isolate the economic, environmental, and reliability impacts of PV self‑consumption and export, diesel backup operation under grid outages, and WHR utilization. To assess the robustness of the proposed framework, sensitivity analyses are performed. Results indicate that under low industrial electricity tariffs, PV primarily contributes to emission reduction, whereas attractive feed-in tariffs transform PV into the most economically beneficial technology through electricity export. WHR consistently reduces natural gas consumption and associated emissions but remains economically attractive only under favorable fuel pricing or investment conditions. Diesel generation becomes economically justified only when grid outages and production losses are considered, highlighting its role as a resilience asset rather than a routine energy source. Overall, the proposed framework provides a transparent, computationally efficient, and policy-relevant decision-support tool for industrial EH planning under evolving market and regulatory conditions. https://eej.aut.ac.ir/article_6131.html This paper presents a high step-up DC–DC converter with enhanced performance characteristics suitable for low-voltage renewable energy applications. The proposed topology achieves high voltage gain over a wide duty-cycle range without imposing excessive electrical stress on semiconductor devices. Owing to its structural configuration and energy transfer mechanism, the converter exhibits low input current ripple and operates effectively in continuous conduction mode. The voltage and current stresses across the switches and diodes are significantly reduced, allowing the use of lower-rated power devices while maintaining high efficiency. The converter also features a common ground between the input and output terminals, which simplifies gate-driver design, reduces electromagnetic interference, and improves overall system safety. Its structure is straightforward and practical for implementation in renewable energy systems such as fuel cells, photovoltaic modules, and electric vehicle power systems. A 240 W laboratory prototype has been developed to validate the theoretical analysis, achieving a peak efficiency of 93.4% at rated load. https://eej.aut.ac.ir/article_6132.html This paper proposes an optimal tuning framework based on artificial neural networks for Laguerre function-based model predictive control using exponential data weighting (ANN-Laguerre-EDW MPC) applied to load frequency control (LFC) in renewable-integrated microgrids. The proposed approach eliminates the need for iterative optimization strategies by utilizing an ANN as a surrogate model to learn the relationship between controller settings and system performance. The approach is evaluated against conventional PID, MPC, Laguerre-EDW MPC, and PSO-based Laguerre-EDW MPC approaches under both step and random load perturbation conditions. Simulation results indicate that the proposed approach achieves up to a 68% reduction in the performance index (ISE) value relative to the other approaches, while facilitating smoother control actions and minimizing frequency oscillations. Moreover, the approach significantly decreases computing time by as much as 95% relative to PSO-based optimization, underscoring its computational efficiency and suitability for practical microgrid control implementation. The proposed LFC approach is further substantiated through sensitivity analysis against parametric uncertainty and statistical analysis, affirming its robustness and reliable performance. https://eej.aut.ac.ir/article_6144.html Hand gesture recognition is a pivotal area in artificial intelligence, enabling seamless human–machine interaction across diverse applications such as home appliances, electronic devices, and public transportation systems. This paper introduces the Moghbeli Persian Sign Language (MPSL) database, a comprehensive image-based resource designed to advance Persian sign language recognition research. The database comprises 107,630 images organized into 47 classes, including 37 Persian alphabet characters and 10 numerical digits (0–9). The images were extracted from 1–2-minute videos captured using a mobile phone camera and involve 18 participants (seven women and eleven men) fluent in Persian sign language. Data acquisition was performed under varying lighting conditions and imaging environments to enhance dataset diversity and robustness. Each class contains 2,290 images, resulting in a balanced large-scale database suitable for training and evaluation purposes. To validate the proposed database, four deep learning and hybrid recognition models were evaluated. The experimental results demonstrate that all evaluated methods achieve high recognition performance on the Moghbeli Persian Sign Language database. Among the evaluated methods, the hybrid model combining Oriented FAST and Rotated BRIEF, Gabor filters, and a Convolutional Neural Network achieved the highest average accuracy of 99.94% while requiring only 241,343 trainable parameters. The obtained results confirm that the proposed database provides sufficiently discriminative visual information for reliable recognition of Persian sign language characters and numerical digits. This database therefore represents a valuable benchmark resource for future research in Persian sign language recognition. https://eej.aut.ac.ir/article_6145.html This paper proposes a dual-band rectangular stub-loaded monopole antenna suitable for Sub - 6 GHz and IoT uses. The proposed antenna employs a rectangular radiating patch with two incorporated rectangular stubs and two rectangular slots to provide wideband impedance matching and dual-frequency operation. The antenna is fabricated on a Roger RO3003 substrate on compact dimensions of 50 × 50 × 1.524 mm³ along with a 50 Ω microstrip feedline to effect good impedance matching. The measured S₁₁ results have shown two distinct operating band: 2.2GHz–2.6 GHz and 3.0GHz–4.9 GHz, with an exceptional return loss of –68.57 dB at 4.22 GHz, exhibiting excellent impedance matching. The suggested antenna exhibits a top gain of 2.72 dBi & total efficiency up to 97.8% that promises very good radiation performance. Due to its small size, high efficiency, and excellent impedance characteristics, the antenna is highly appropriate for IoT, WLAN, and Sub-6 GHz wireless communication systems. https://eej.aut.ac.ir/article_6146.html The surge in use and access to the Internet of Things (IoT) devices has created many challenges for securing data communications particularly when they interface with cloud-based infrastructures. Traditional means of encrypting data are inadequate due to the limited resources of IoT nodes and the constantly changing nature of cyber threats; especially with the potential emergence of quantum computers. To overcome these security challenges, a hybrid cryptographic framework has been developed (chaotic Advanced Encryption Standard (AES-256) /Elliptic Curve Cryptography (ECC) with Puma encryption), which will also include quantum-resilient cryptographic principles to ensure secure key creation and exchange regardless of future quantum threats. An important part of this approach is the ability to generate secure cryptographic keys using a chaotic Puma optimization approach. The core cryptographic algorithm is the asymmetric AES-256 encryption algorithm which provides fast, Block-Based Data Encryption while the Asymmetric Cryptography handling secure keys exchange and digital signature validation for preventing Man-In-The-Middle (MITM) and Spoofing Attacks on Keys. Many assessments are performed on this model using java with metrics such as Encryption Time, Key Setup Time, Memory Overhead, Latency, and Error Propagation Rate. Test data will consist of same structured data from the Standard Dew Point Method and other sources, along with Real-World Sensor Emulation data. The results indicate that the proposed architecture has significant advantages over other solutions (PQC and DH) in terms of computational overhead, latency (minimal), and ability to resist data corruption over lossy channels. Furthermore, ECC’s small key sizes make it suitable for use to secure resource-constrained edge nodes without sacrificing strength. This research demonstrates that a layered cryptographic approach is both practical and offers a forward-compatible path to quantum-resistant systems for robust, scalable IoT-cloud security. The findings contribute to creating lightweight, secure, and resilient communication frameworks for the next generation of Cyber-Physical and IoT environments. https://eej.aut.ac.ir/article_6147.html In this paper, touch and step voltages of different arrangements of rods including single rod, double, triple and quadruple rods under lightning strikes are evaluated. In all analyses, an efficient modeling approach called improved multi-conductor transmission line is adopted. Also, two typical lightning currents namely first and subsequent stroke currents are injected to the mentioned grounding rods. This paper consists of two parts. In the first part, the touch and step voltages of different arrangements of rods considering dispersion and ionization of lossy earth are computed. Firstly, sensitivity analyses show that from single rod to quadruple rods, both voltages are reduced especially in poorly resistive earths. The relative reduction percentage is on average 81.5% and 69.5% respectively for the touch and step voltages. In addition, when the arm length is increased, considerable reduction of both voltages especially in poorly resistive earths under subsequent stroke current is achieved so that the length variation from 10m to 40m, leads to the reduction of 85% and 91% respectively for the touch and step voltages on average. Depending on the earth resistivity and lightning waveform, after a starting length this reduction is slightly continued. In the second part, the above analyses are extended to two-layer earth and sensitivity analysis versus the resistivity of the two layers is carried out. The simulation results show that both voltages have middle values between the counterparts in single-layer earths. https://eej.aut.ac.ir/article_6155.html Matrix multiplication is a fundamental operation in image processing and many data-intensive applications, and its growing computational and memory demands increasingly challenge conventional CPU- and GPU-based systems. Processing-near/in-Memory (PnM/PiM) architectures address this challenge by reducing data movement between memory and computation units. UPMEM is a commercially available DRAM-based PIM platform that integrates lightweight processors within memory chips; however, efficiently mapping compute-intensive kernels such as dense matrix multiplication onto this architecture remains non-trivial due to limited arithmetic support, restricted on-chip memory, and low per-core frequency. In this work, we investigate the feasibility and performance of dense matrix multiplication on UPMEM. Starting from naive implementation, we apply loop reordering and tiling optimizations tailored to the UPMEM execution model. We analyze the impact of key architectural parameters, including the number of DPUs and tasklets, and compare UPMEM-based implementations against optimized CPU baselines under identical input sizes and precision constraints. Experimental results show that exploiting tasklet-level parallelism and DPU scalability yields significant performance improvements, with the tiling-based UPMEM implementation achieving up to 2.76x speedup over CPU tiling and outperforming loop-reordered CPU implementations for large matrices using 32-bit arithmetic. https://eej.aut.ac.ir/article_6174.html As solar-powered electric vehicles (EVs) become more prevalent, intelligent energy management techniques are necessary to optimise the use of renewable energy and boost system efficiency. For sustainable EV operation, this paper proposes an AI-enabled energy management architecture that combines an optimised maximum power point tracking (MPPT) controller, a Coupled Inductor SEPIC Clamp (CISC) converter and intelligent route prediction. By diminishing input current ripple, lowering switch voltage stress and increasing conversion efficiency, the CISC converter is employed for increasing the Photovoltaic (PV) voltage. A Superb Fairy Optimization-based Radial Basis Function Inference System (SFO-RBFIS) MPPT controller is developed for optimising solar energy extraction under dynamic irradiance and temperature circumstances. It offers quick convergence, excellent tracking accuracy and few steady-state oscillations. Furthermore, an Artificial Neural Network (ANN) and the SFO Algorithm are employed in a routing strategy that anticipates energy-efficient routing. The developed system attains a converter efficiency of 97.3%, greater MPPT performance and superior energy management when compared to traditional approaches, according to outcomes in the MATLAB tool. The developed framework offers a sustainable and energy-efficient solution for future solar-powered EV systems. https://eej.aut.ac.ir/article_6175.html Electric Vehicles (EV’s) are measured as an excellent replacement for fuel based vehicle due to their limited emission of greenhouse gases. Despite that, it is essential to establish better EV charging system to enlarge EV utilization. Thus, this paper proposes an enhanced EV charging process based on dual PhotoVoltaic (PV) using High Gain Luo converter with optimized PI controller. High Gain Luo Converter is deployed for both PV system for boosting their output power supply, which enables to achieve higher efficiency with improved Voltage gain, power quality with reduced ripples. For attaining enhanced converter performance, Falcon Optimized Algorithm (FAO) based PI controller is utilized which enables faster response rate with reduced steady-state errors with higher quality results. Furthermore, Bidirectional DC-DC (BDC) converter is utilized to attain overall charging and discharging level of EV battery where separate ANN controllers are employed for monitoring and controlling the charging operation as well as bidirectional converter. Moreover, to determine performance proficiency of proposed system, MATLAB simulation is implemented, from which it is remarkable that dual PV powered EV charging system attained higher efficiency of (95.73%), Voltage gain and reduced THD of (2.87). Therefore, this system ensures superior EV charging with continuous power supply and accurate detection of EV charging level. https://eej.aut.ac.ir/article_6176.html This paper introduces a non-isolated single-input dual-output (SIDO) multiport DC-DC converter with applications in renewable energy systems, electric vehicles, uninterruptible power supplies, battery chargers, and personal computers. The proposed SIDO converter consists of two independent outputs, regulated at 110 V (as a boost stage) and 12 V (as a buck stage), respectively. Key features include an optimal 12-component count, continuous input currents, a common-ground architecture, and independent controllability of output voltages. Furthermore, the converter achieves the desired voltage gain, low voltage stress on diodes and switches, and a high efficiency of approximately 93.1%. The system is designed for a 245.77 W total power rating, providing 132.5 W for the boost stage and 113.3 W for the buck stage. The proposed topology is tested in the laboratory, and the experimental results verify the proper performance of the converter. https://eej.aut.ac.ir/article_6177.html The human brain is a complex, integrated system defined by higher-order, multi-way interactions that are inherently simplified by traditional pairwise network models. Although a diverse array of analytical methods has been developed to address this limitation, the literature remains fragmented, lacking a unifying conceptual framework to organize these emerging developments. This review provides a comprehensive synthesis of higher-order functional connectivity (HOFC) methodologies in brain connectomics. We propose a fundamental distinction between two complementary paradigms based on their analytical outputs: Topographical HOFC and Temporal HOFC. We trace the evolution of Topographical HOFC from foundational Correlation-of-Correlations (CoC) and spatial profile similarity measures to advanced multi-view fusion frameworks and graph neural networks. Concurrently, we survey Temporal HOFC, exploring dynamic sliding-window clustering, meta-connectivity, high-resolution edge-centric models, and statistically rigorous, model-based approaches like the Matrix Variate Normal Distribution (MVND). By critically examining the conceptual, statistical, and computational challenges of these frameworks—including the biological interpretation gap, the "genuineness" of observed interactions, and the curse of dimensionality—we outline key comparative insights. Ultimately, we argue that the future of the field lies in a principled, neurobiologically grounded integration of these spatial and temporal approaches to unlock a robust, multi-level understanding of complex brain organization in health and disease. https://eej.aut.ac.ir/article_6178.html Rising electricity demand causes mismatches between generation and demand, leading to frequency deviations from the nominal 50 Hz. Ethiopia's hydropower plants rely on primary load frequency control mechanisms; however, a supplementary mechanism is needed to maintain frequency centrally. This paper presents a hybrid neuro-fuzzy inference system-based fuzzy-proportional-integral-derivative controller and compares it with a fuzzy-proportional-integral-derivative controller, a neuro-fuzzy inference system-based controller, and a conventional proportional-integral-derivative controller within a three-area load frequency control model. The aim is to provide a robust, supplementary, centralized mechanism for the Ethiopian Electric Power grid that resolves challenges in real-time controller-parameter tuning. The hydro-governor, hydraulic turbine, synchronous generator, tie-line, and load were modeled, and a three-area system model was formulated. All four controllers were designed in MATLAB/Simulink using three governor models, and their performance was evaluated comparatively. Frequency and tie-line power flow changes, which determine the area control error, serve as feedback to the governor for regulating turbine water flow. Response was assessed via overshoot, undershoot, steady-state error, and settling time. The proposed hybrid controller significantly outperforms the alternatives: in Scenario 1, settling time fell to 12.959 seconds with an integral time absolute error of 0.00565, versus 48.8824 seconds and 0.0838 for the conventional controller. Under multi-area disturbances, the hybrid controller retained the lowest error of 0.02375, and robustness analysis under plus/minus twenty-five percent parameter variations confirmed stability. The novelty lies in hybridizing a Mamdani fuzzy-gain scheduler with a Sugeno-type neuro-fuzzy system trained on multi-governor transient data, a combination not previously reported. https://eej.aut.ac.ir/article_6179.html Precise identification of Fetal movements is necessary for continuous prenatal monitoring; however, the signals obtained from wearable accelerometers are always affected by artifacts induced by the maternal movements and environmental noises. This research proposes a hierarchical signal refinement paradigm that can iteratively refine inertial sensor signal data from wearables in order to detect fetal movements in realistic monitoring environments. The hierarchical signal refinement paradigm uses a well-defined process to eliminate noise, extract compact features, and extract discriminative features to be able to conduct multiclass classification. Performance evaluation of this approach has been conducted using a publicly available fetal movements benchmark dataset containing synchronized tri-axial accelerometer data corresponding to fetal movements, mother's respiration, and laughing. The experimental results showed 97.94% of accuracy, 97.00% of precision, 96.00% of recall, 97.19% of F1-score, and 98.21% of area under the receiver operating characteristic curve. In particular, the ablation experiment validated the contributions of each signal refinement step, while statistical significance and computational cost analysis further confirmed the effectiveness, reliability, and efficiency of the proposed framework. Compared with recent fetal movement recognition methods, HSRF provides an interpretable and effective signal refinement strategy suitable for continuous wearable prenatal monitoring and intelligent maternal healthcare applications. https://eej.aut.ac.ir/article_6180.html Indoor localization is a key enabler of smart environments, yet achieving high-accuracy positioning remains difficult due to severe multipath effects, signal distortion, and the limited ability of current deep models to generalize across heterogeneous wireless datasets. Most existing approaches overlook the intrinsic sparsity structure of wireless signal features, leading to unstable latent representations and reduced robustness in complex indoor scenarios. To bridge this gap, this paper proposes a non-smooth regularized autoencoder that embeds a sparsity-inducing non-smooth regularizer into the latent space, enabling the extraction of compact, noise-resilient, and interpretable spatial features from high-dimensional signal measurements. This design introduces a clear advancement over state-of-the-art autoencoder and regression-based models, which predominantly rely on smooth penalties and consequently struggle under strong non line of sight conditions. Evaluations on the UJIIndoorLoc benchmark show that the non-smooth regularized autoencoder achieves an average localization error of 6.78 m, outperforming conventional deep neural network baselines. Furthermore, experiments on a real-world CSI dataset collected in a 13.5 × 11 m² cluttered laboratory demonstrate a mean error of 1.73 m in a challenging non line of sight environment. These findings confirm that the proposed non-smooth regularized autoencoder effectively addresses core challenges in robust feature learning and offers strong generalization across diverse indoor settings, establishing it as a powerful framework for next-generation indoor localization systems. https://eej.aut.ac.ir/article_6181.html This paper deals with the design of an effective smart distributed management system (DMS) in islanded microgrid connected with multiple nanogrids under renewable energy variability using Fuzzy Logic method. The proposed DMS strategy was coordinated with an optimal Under-frequency Load Shedding (UFLS) to ensure power equilibrium during overloads. The recently nature-inspired named African Vultures Optimization (AVO) algorithm was employed for this purpose. The investigated microgrid includes doubly-fed induction generators (DFIG) wind farms, solar PV generators and concentrating solar power (CSP) units. Further, a storage system including Redox Flow Batteries (RFB) was used to support frequency regulation. Further, to show the validity and potential of the proposed method a variety of nanogrid systems were tested involving the sea-water desalination station, pumped solar system and electrical vehicles station. Furthermore, the proposed smart DMS coordinated with intelligent UFLS can maintain the microgrid stability using the smart deloading of the connected nanogrids using an estimated load duration curve (LDC) which minimizes the outage risk and avoids the total load shedding during faults. Using this approach, frequency stability can be maintained after disturbances. The reached results show good performances in view of frequency peak under/overshoot and settling time.