ABSTRACT. This work proposes a methodology for studying the behavior of corona discharge currents using a laboratory scale hybrid AC-DC rod-plane experimental arrangement operating at reduced voltage levels. This methodology enables emulation of surface electric field gradients encountered on high voltage conductors without applying high voltage. A rod-plane system was designed and constructed to measure conducted emissions via a shunt resistor and radiated emissions in the frequency range of 0.1 to 1,000 MHz.

ABSTRACT. This paper presents a continuous under-frequency load shedding (UFLS) scheme based on the rate of change of frequency (ROCOF), applied to the IEEE 39-bus system. Unlike conventional UFLS schemes, which operate in fixed-frequency threshold stages, the proposed method enables proportional and continuous load disconnection according to the severity of the disturbance. The control logic is implemented externally using Python, directly linked with dynamic simulations performed in PSS®E. As a case study, the simultaneous loss of generators at buses 32 and 34 is analyzed, replicating a critical contingency identified in previous research. The results show that the proposed scheme significantly improves the frequency nadir, reduces the amount of load shed, and prevents system collapse, outperforming conventional strategies. It is concluded that the continuous ROCOF-based scheme offers an efficient, adaptive, and fast-responding solution to enhance frequency stability in modern power systems.

ABSTRACT. This paper presents a novel state-feedback optimal control design for frequency and voltage regulation in power systems. Typically the control methods used in power systems are based on the system outputs, such as proportional-integral derivative controllers and lead–lag compensators; nonetheless, those methods usually do not take advantages of the model characteristics, and their performance can be very sensitive to the controller tuning and the system setting. On the other hand, a model-based controller can considerably improve the system performance by modifying in an effective way its dynamics as desired or required. A novel tuning controller methodology is introduced by using the modal participation factor analysis for identifying those variables related to the poor damped modes and then modify its behavior to achieve an effective, faster and smoother system response. Fur such a purpose, based on the power system model, the optimal controller is designed, resulting in a state-feedback control law, which inherently improves the system performance and robustness. By its part, the controller requires to be tuned, usually by a trial and error method, through its corresponding parameters, known as weighting matrices, and it is here where our proposal contributes by proposing a formal modal-based controller tuning method. The proposed strategy is applied to the IEEE 9-bus system, evidencing the controller capabilities in achieving an outstanding system performance in the simulation results.

ABSTRACT. This paper presents the control methodology for a three-phase permanent magnet synchronous motor. By considering that the motor has two control inputs in the dq reference frame, then two outputs are selected to be robustly regulated: the velocity and the stator current to achieve maximum electromagnetic torque operation. Firstly, the motor dynamics is modeled such that a feedback linearization can be developed, then the robust controller is designed. Secondly, the electromagnetic torque is maximized through the corresponding control input. The effectiveness and robustness of the proposed control methodology are evaluated by considering abrupt changes in the disturbance (load torque) and sudden velocity changes. The control system’s performance is validated with numerical simulations.

ABSTRACT. Abstract—Nonlinear loads cause harmonic distortion and low power factor in industrial systems. The resulting flow of harmonic currents alters the ac power signals, leading to the malfunctioning of electrical and electronic equipment. Capacitor banks and tuned passive filters have been used to improve power factor (PF) and reduce harmonic currents, respectively. However, industry remains concerned about the vulnerability of these methods to harmonic resonances and voltage distortion, which could damage valuable components. Active power filters (APFs) have been developed over recent years and are now a practical solution for these issues. In this paper, a selective harmonic compensation APF is proposed for harmonic mitigation and PF enhancement in industrial systems. The APF employs the Synchronous Reference Frame (SRF) technique to detect and reduce individual or sets of harmonic currents using d and q variables. To provide a comprehensive performance comparison, single-tuned passive filters are also designed and simulated for the same harmonic spectrum. Both solutions are evaluated under identical load and grid conditions to analyze their effectiveness in harmonic mitigation and power factor correction. Simulation results are presented to assess the proposed solution and demonstrate correct APF operation.

ABSTRACT. Traditional methods of teaching automatic control often lack practical involvement, making it difficult for under-graduate engineering students to connect theory with real-world applications. This work presents the development of a low-cost pedagogical prototype aimed at teaching automatic control principles through hands-on experimentation. The prototype integrates a direct current (DC) motor, an encoder, a current sensor, a DC motor driver, and an embedded system. In addition, a Graphical User Interface (GUI) was developed in the MATLAB environment. The GUI communicates through the serial port of the PC and allows students to define the experimental parameters and monitor responses in real time. The system enables students to design and apply open-loop and closed-loop controllers in a practical environment, enhancing their understanding of control theory. The low-cost and user-friendly design of this prototype makes it accessible for educational purposes, bridging the gap between theoretical concepts and practical applications.

ABSTRACT. This paper addresses the problem of fault diagnosis in boost converters through the design and simulation of observers based on convex Takagi-Sugeno (T-S) models. This type of converters are used in power systems, making them susceptible to faults that can compromise system stability and efficiency. A fault diagnosis scheme is proposed that utilizes the convex Takagi-Sugeno model to represent the nonlinear behavior of the converter, as well as a convex T-S observer to generate the residuals, which function as fault indicators. The simulation results show that the proposed approach enables early and accurate fault detection, which contributes to improving the operational reliability of the system.

ABSTRACT. This paper presents a comparative analysis between the classical Takagi-Sugeno (TS) convex modeling approach and a novel fuzzy perturbation (FP) method applied to a nonlinear DC motor with an eccentric load. The FP approach separates the nonlinearities as a perturbation, allowing conventional linear control techniques to be used. Both modeling strategies are simulated and compared against the original nonlinear model. Two different control objectives are addressed: position tracking and speed regulation. For the TS model, a parallel distributed compensator is designed using Linear Matrix Inequalities (LMIs), while the FP model uses state feedback combined with a fuzzy feedforward strategy. Results show that while both methods perform adequately for position tracking, the FP model significantly outperforms the TS approach in speed control, especially under large references, offering better scalability and reduced risk of actuator saturation. The proposed FP model proves to be a promising alternative for fuzzy modeling and control in nonlinear systems.

ABSTRACT. This paper presents the design and implementation of a passivity-based control (PBC) strategy for DC-DC boost converters, enhanced with a novel fuzzy logic-based self-tuning mechanism for damping injection. The proposed approach pre- serves the structural and energetic properties of classical PBC, while addressing a critical practical limitation: the manual tuning of damping-related parameters, such as parasitic inductor resis- tance and load-associated resistance. To overcome this challenge, an adaptive scheme is introduced that dynamically adjusts these parameters in real-time based on system behavior, using fuzzy logic rules driven by state error signals. This adaptive capability enables robust performance under varying operating conditions and unmodeled disturbances, thus significantly improving the practicality and reliability of PBC in real-world applications. The findings of this study represent a significant contribution to the field of nonlinear control for power converters, paving the way for the development of more intelligent and resilient energy conversion systems.

ABSTRACT. This paper presents the design and experimental development of a high-efficiency lithium-ion battery charger, which includes a pack of 24 INR18650-35E cells arranged in 4 branches of 6 cells in series each, intended for electric mobility and energy storage applications. The proposed charger incorporates a power conversion stage based on a medium-frequency isolated topology, enabling a compact and lightweight design with enhanced performance. The system employs state-of-the-art silicon carbide (SiC) power electronics components, allowing high switching frequencies and improved thermal behavior. Experimental validation of a 300 W, 22.2 V nominal prototype demonstrates a peak efficiency of 92.3% at full load and 98.14% efficiency at the transformer terminals. The proposed solution addresses the growing demand for efficient, reliable, and scalable charging systems for modern lithium-ion battery applications.

ABSTRACT. Autonomous vehicles use different sensors to perceive their surroundings and make driving decisions. Among these, LiDAR is probably the most useful sensor due to its high accuracy and role in supporting various autonomous driving tasks. However, despite significant advances in autonomous driving technology, multiple reports of attacks targeting the sensors of these vehicles have been documented. In particular, LiDAR point insertion attacks have drawn increasing attention from the research community, as they can cause vehicles to brake abruptly or even result in fatal collisions. Although several countermeasures have been proposed, many fail to provide comprehensive protection, often exhibiting vulnerabilities that hinder their effectiveness in real-world scenarios. This work presents a watermarking scheme to address LiDAR point insertion attacks. Our approach segments the LiDAR point cloud into voxels and embeds a watermark message in those enclosing points. The robustness of the proposed method is evaluated against such attacks using points collected from real-world LiDAR readings.

ABSTRACT. Photovoltaic systems have emerged as one of the most effective solutions to meet energy demands while preserving the environment. However, the performance of these systems can be compromised by a range of faults, including partial shading, soiling, disconnections, module defects, and degradation of electronic components. This article proposes a data-driven approach based on topological data analysis for fault diagnosis in the power analysis of PV systems, aiming to contribute to more robust and efficient system monitoring. The proposed methodology employs the bottleneck distance as a comparative metric, enabling the detection of behavioral deviations as small as 5%. The simulated scenarios show that the proposed methodology is sensitive enough to detect subtle changes, such as the removal of a single panel or small losses due to environmental factors.

ABSTRACT. Zapoteco is an Indigenous language still spoken in southern Mexico, particularly in the state of Oaxaca, where it has many regional variants. Although the language persists, the dominance of Spanish as the primary language threatens its survival, with the risk of decline in the coming years. To preserve Zapoteco culture and its associated traditions, technological strategies are being explored. Artificial Intelligence, particularly through Natural Language Processing (NLP) models, offers a promising solution by developing tailored translation systems. These tools would assist Zapoteco speakers who have limited proficiency in Spanish, enabling them to carry out daily activities in their native language. This study explores how three machine translation models—mBART, MarianMT, and M2M100—can be adapted to Zapoteco, an Indigenous language from Oaxaca with very few available digital resources. Given the lack of data, synthetic sentences inspired by the everyday language of rural communities were created, allowing the models to be fine-tuned for this specific task. The results revealed that mBART delivered the best translations, followed by M2M100, while MarianMT showed more limitations in contextualizing. The research highlights that, with creative strategies such as synthetic data generation, it is possible to significantly improve translation for underdocumented Indigenous languages.

ABSTRACT. In recent years, intelligent video surveillance systems have become a broad field of research. A video surveillance system consists of a camera network that observes a specific region of space. These cameras may or may not share the observation region (overlapping or non-overlapping). The collection of cameras may have different resolutions and characteristics, which complicates the study of video sequences due to several issues, such as the acquisition distance and the very low quality of images. Among the main problems addressed in video surveillance systems is the person re-identification problem, which is about identifying the same person across a set of connected cameras; these can be non-overlapped or overlapped. In this work, the advances of Deep Learning models are considered, specifically Transformers-based, trying to get a better image representation to improve the person re-identification problem results. We used several Transformers models to represent images of people, and to test them, we used the popular CAVIAR4REID dataset. The results were obtained with several models, with some interesting results and conclusions.

ABSTRACT. The explicability is the ability that Artificial Intelligence models must have to be understood in some way, thereby enhancing trust in their results and enabling the use of these models to make more informed decisions. Specifically, in the natural language processing area, this trait is becoming more essential, since the explosion of the usage of Large Language Models in any academic, labor, or governmental area, it is necessary to increase the level of trust in these models by the final users and decision makers.

Taking this into account, this work evaluates three text classification techniques using a user profiling dataset with two tasks, gender and nationality classification problems. In this case, the dataset has users' tweets in Spanish from several nationalities. Furthermore, we explore some explainability traits using the model's outputs, specifically weighted schemes. From that information, we build word clouds representing the tokens (or words) with the highest score, which have the most relevance in the classification performance.

Interesting findings are discussed in the conclusions, but we obtained some explainability with high performance in these two classification problems.

ABSTRACT. Agent-based models (ABMs) are widely used to simulate complex biological systems at single-cell resolution. However, their high computational cost limits their applicability to large-scale or high-throughput studies. In this work, we present an off-lattice version of the Cess–Finley tumor–immune model using the PhysiCell framework. The model incorporates extended immune signaling rules, capturing IL‑10 and hypoxia effects, and runs in a center-based, off-lattice environment with OpenMP-based parallelism. We analyze scalability through strong and weak scaling experiments, achieving a maximum speedup of 6.1x on 8 cores and maintaining 59% efficiency in weak scaling. To assess real-world performance, we conducted a Multi-Parameter Sensitivity Analysis (MPSA) comprising 7,600 simulations. The proposed parallel algorithm enables biologically detailed simulations with improved computational efficiency.

ABSTRACT. Arboviral diseases have a notable impact worldwide, particularly in most Latin America. Cases are on the rise in frequency and magnitude, making them an important public health problem with social and economic consequences. For this reason, multiple efforts exist to combat these diseases; one of the main challenges is the design of anti-arboviral drugs, which do not currently exist. Peptides have been successful in the last decades in treating different antiviral diseases, so the discovery of peptides with anti-arboviral therapeutic potential acquires great relevance. During this period, different machine learning approaches have been used for this purpose. Therefore, we consider applying classical models using a database proposed in this work from the manual compilation of the data available in the literature. The results obtained are promising, constituting an interesting starting point for deepening this novel problem.

ABSTRACT. Chronic kidney disease represents a significant global health challenge, characterized by the progressive deterioration of renal function associated with elevated morbidity and mortality. Early identification is essential for preventing adverse outcomes and improving clinical management. This study aimed to evaluate and compare tree-based machine learning models for classifying of chronic kidney disease, with an emphasis on predictive performance and model interpretability. Four decision tree-based classifiers were implemented and evaluated using stratified train-test splits. Missing values were imputed using the k-nearest neighbors algorithm, and class imbalance was addressed through the application of the synthetic minority over-sampling technique. Among the models, CatBoost achieved the highest performance, reaching an area under the curve of 0.952, along with a sensitivity of 0.824 and specificity of 0.902, indicating a robust ability to classify both positive and negative cases correctly. SHapley Additive exPlanations analysis was conducted to interpret model predictions, identifying time to event months, HgbA1C, and eGRF baseline as the most influential features. These findings support the utility of CatBoost for severity classification in chronic kidney disease, particularly due to its effective handling of categorical variables and reliable predictive capabilities.

ABSTRACT. Periodontitis and gingivitis are highly prevalent conditions that affect both oral and systemic health. This work presents an automatic classification methodology based on artificial neural networks (ANNs) to distinguish between healthy, gingivitis, and periodontitis conditions using a dataset that integrates clinical variables and intraoral infrared thermographic features. Z-score normalization and dimensionality reduction via Principal Component Analysis (PCA) were applied to optimize training and enhance model generalization. Multiple network architectures were tested, ranging from deep models to single-layer perceptrons. The best performance was achieved with a multilayer perceptron (MLP) consisting of one hidden layer with 11 neurons and sigmoid activation, reaching a classification accuracy of 84.56%. Results indicate that with proper feature representation, simpler models can outperform more complex ones, offering a practical and interpretable solution for embedded diagnostic systems or mobile clinical support tools.

ABSTRACT. This paper presents the design and implementation of an off-grid photovoltaic (PV) system integrated with battery energy storage, focusing on energy management and stability control in MATLAB/Simulink. The proposed system combines a 100 kWp PV array with lithium-ion batteries managed through bidirectional power converters and a hierarchical control strategy. The design includes DC-DC boost converters with Maximum Power Point Tracking (MPPT) based on the Perturb & Observe algorithm, buck-boost converters for battery charge/discharge control, and a three-phase inverter with LCL filtering for AC conversion. The control system maintains voltage and frequency stability under varying load conditions (50-100 kW) and irradiance levels (500-1000 W/m²). Simulation results demonstrate the system's ability to handle rapid transitions between energy surplus and deficit conditions while ensuring power quality. The proposed configuration offers a reliable solution for off-grid applications, particularly in remote areas without access to conventional power grids. The study provides practical insights into the integration of renewable generation with energy storage systems, addressing key challenges in standalone microgrid operation

ABSTRACT. This paper introduces the Circular Energy Sustainability Index (CESI), a novel tool designed to evaluate renewable energy technologies through the lens of the circular economy. The index incorporates key factors such as design circularity, resource efficiency, and waste recovery. When applied to the most widely used renewable energy technologies, namely photovoltaic and wind systems, the CESI reveals significant opportunities to enhance the circularity and sustainability of the energy sector. The results indicate that photovoltaic energy has an 82% sustainability rating, while wind energy has a 49% rating. This indicates that wind energy has more critical areas to address, as its composite materials are more difficult to recycle. Furthermore, photovoltaic systems are more circular, while there is still room for further optimization in the selection of materials to improve their sustainability rating. These insights support a more resilient and equitable energy transition. The index provides a rigorous and comprehensive environmental assessment, identifying the most sustainable technologies and highlighting areas for improvement to reduce the environmental impact of the energy sector.

ABSTRACT. This research introduces a comprehensive optimization framework for hybrid microgrid energy management that integrates battery storage, wind, and solar PV under bounded uncertainty. The method combats the inherent uncertainties in renewable generation and load demand by integrating robust optimization and dynamic programming. Using real data, it establishes realistic uncertainty sets for solar irradiance, wind speed, and electrical demand, enabling optimal decision-making in the face of extreme fluctuations without the need for computationally intensive stochastic simulations. The control strategy reduces a comprehensive cost function that encompasses grid energy, transmission losses, load deficits, and battery degradation while adhering to all operational constraints. In comparison to conventional methods, numerical results confirm that this dynamic programming framework is significantly more cost effective and reliable, demonstrating a scalable and resilient solution for sustainable microgrid operation in variable conditions.

ABSTRACT. A methodology to control a Doubly-Fed Induction Generator (DFIG) in zone III without a pitch controller is described. In zone III, the mechanical power harvested through the wind kinetic energy exceeds the rated generator power. Thus, when operating in this zone, the control objective is to keep the generator’s rated power constant. The classical asynchronous machine model in per unit is adapted to satisfy generator operation, power delivered to the grid at the stator winding while preserving the Port-Controlled Hamiltonian (PCH) structure. A control law is synthesized using energy-shaping control by varying the rotor angular velocity reference according to zone III demands. A series of simulations keeping the pitch angle constant are presented in order to show the effectiveness of the proposed methodology.

ABSTRACT. Innovation in the development of power electronics is evolving on a large scale, offering greater presence of Inverter-Based Resources (IBRs) in the electrical power system. However, this also gives rise to new challenges, such as maintaining the operational integrity of the system in the event of faults in transmission lines with high IBR penetration. In this sense, traditional protection elements may not operate properly and consequently lead the system to undesirable operating conditions. Therefore, this paper quantitatively evaluates the impact of inverter-based resources on the accuracy of one- and two-terminal fault location methods in different scenarios and with different levels of renewable energy penetration in the 39-node IEEE system. Both methods were compared based on the relative error of the fault distance, finding that the single-terminal method increases the error rate compared to the two-terminal method, which exhibits greater robustness.

ABSTRACT. Nowadays, the renewable energy networks are designed for transmission and distribution using underground systems. Such systems minimize visual impact, reduce congestion, and increase security and reliability. However, the technology utilized has increased the capacitance effect, which can lead to the system being more susceptible to the occurrence of the ferroresonance phenomenon. Ferroresonance can cause dielectric or thermal damage and consequently makes the system susceptible to faults in the electrical power equipment and installations. This paper analyzes the numerical solution of an RLC circuit considering the nonlinearity of the inductor through a hyperbolic sine function, validating the data obtained with a simulation of voltage, current and magnetic flux. Subsequently, the ferroresonance phenomenon in photovoltaic parks is analyzed to detect possible ferroresonance conditions.

ABSTRACT. The focus of energy production in the last decade has been oriented towards the generation of green energies, such as wind energy. As a result, there has been a big development in a variety of arrays that implement different types of generators and control systems, in which it is important to consider the climate variations that can, and will, affect the electrical power delivered by the generators. Because of said intermittence, the more practical solution is the utilization of generation systems that allow the variation of wind speed and avoid the problems that come with it, that is, the implementation of electrical machines that work under these changing wind speed conditions without causing any synchronism issues with the electrical grid. Where the usage of wound induction generators allows for a wide control of its generation variables. In this article, the controller for a Doubly Fed Induction Generator (DFIG) is modeled, which enables the control of the electromechanical generation variables efficiently. Given the importance of the DFIG parameters, having an appropriate control of the active and reactive power achieves a reliable and efficient system. Simulation results are included as to verify the correct operation of the modeled controller under the mentioned wind speed conditions.

ABSTRACT. Large Language Models (LLMs) have transformed natural language processing and are now being adapted for non-textual tasks, including time series forecasting. This paper explores the fine-tuning of LLM-based and LLM-inspired Time Series Foundation Models (TSFMs) for short-term electric load prediction in a regional context. Using historical load data from the Alberta Electric System Operator (AESO) spanning 2011–2023, we evaluate five recent TSFMs: MOMENT, PatchTST, Tiny Time Mixers (TTMixer), TimeLLM, and AutoTimes. While TimeLLM and AutoTimes explicitly repurpose LLMs for time-series modeling, models like MOMENT and PatchTST apply similar transformer-based architectures. Our findings show that out-of-the-box TSFMs underperform due to the complex and evolving nature of Alberta’s electricity consumption, but fine-tuning yields significant improvements. Our results show that medium-sized models, such as TTMixer and TimeLLM outperform larger architectures, achieving high forecasting accuracy and robust uncertainty estimation.

ABSTRACT. The integration of renewable energy sources into power systems requires efficient and reliable energy storage solutions to manage supply fluctuations and ensure grid stability. Supercapacitors offer a compelling option due to their electrostatic energy storage, with long life cycle, and rapid chargedischarge capabilities. This work proposes a Supercapacitorbased Energy Storage System (SCESS) interfaced with the electrical grid via an Active Neutral-Point Clamped (ANPC) converter. The supercapacitor is directly connected to the DC-link without requiring an intermediate converter. The overall system dynamically operates in either two-level or three-level mode based on the supercapacitor voltage, optimizing efficiency. A vector control strategy in the d-q reference frame enables precise active and reactive power management, using PI controllers and a phase-locked loop for synchronization. Simulations demonstrate accurate tracking of power references under various operating conditions. The study highlights the ANPC-based SCESS as a viable and sustainable solution for future smart grids, offering superior cost and efficiency compared to conventional storage systems.

ABSTRACT. In this work, an electromechanical model was developed for electric vehicles (EVs) with lithium-ion batteries (BILs), integrating the dynamics of the electric motor, transmission, and battery discharge. A proportional-integral (PI) controller was used to ensure accurate speed tracking, minimizing errors in the face of disturbances such as gradient changes or aerodynamic drag. The model considers battery energy efficiency and system thermal limitations. The methodology included simulation of the model in a dynamic environment, evaluating controller performance under a driving cycle limited to 1000 seconds. The results show that the PI controller achieves a tracking error of less than 2% under nominal conditions.

ABSTRACT. Introduction: Position estimation systems are essential in fields such as navigation, medicine, agriculture, and construction. Currently, GNSS is one of the most widely used positioning technologies; however, its signal quality significantly deteriorates in enclosed or obstructed environments. Visual odometry (VO) has emerged as a promising alternative, enabling position estimation even in GNSS-denied settings. Nevertheless, one of its main limitations lies in its reduced performance under low-light conditions. A potential solution involves the use of near-infrared (NIR) cameras in combination with optical markers to enhance visibility and accuracy. Material: A literature review was conducted using the Dimensions database to identify studies related to visual odometry, NIR cameras, and optical markers. The inclusion criterion required each selected study to address at least two of these three components. Results: Out of an initial set of 4,334 records, 21 studies met the inclusion criteria. The analysis revealed that optical markers significantly improve position estimation in low-light environments. Furthermore, while the integration of NIR cameras in visual odometry remains limited, the findings suggest it holds great potential for improving system robustness under challenging lighting conditions. Conclusion: Visual odometry systems using optical markers in NIR imaging represent a viable and promising solution for position estimation in complex and GNSS-denied environments.

ABSTRACT. Beekeeping in Mexico has prevailed in the first places in production, occupying sixth place, below India and Ukraine, the challenge is to conserve the Apis mellifera species, it is a situation that concerns everyone, currently the monitoring of hives is face-to-face and the review periods depend directly on the number of apiaries that the producer manages. causing the hive to be at risk for prolonged periods without supervision, in addition to being limited to the observation method and traditional procedures. This paper describes an apiary monitoring system using noninvasive sensors to measure weight and calculate production, helping the beekeeper to detect anomalies within the hive. The acquisition of data allowed one to measure the weight of the apiary in real time, emulating field situation exercises, calculating, in some cases, up to R^2 of 94. 07%, to measure effectiveness. The main conclusions indicate that the system is more efficient than traditional techniques, suggesting that data acquisition to monitor the hive in risk situations is feasible with the use of external sensors.

ABSTRACT. In recent years, the rise of renewable energy has driven the development of new power electronics topologies. Simulation, particularly Hardware-In-the-Loop (HIL), has become essential for testing control algorithms under conditions close to real hardware. However, commercial HIL systems are expensive, and custom solutions require significant time and technical expertise. This work presents a semi-custom HIL solution for a three-phase ANPC inverter using LabVIEW FPGA and the CRIO-9067 platform, achieving a 1 μs computation time and less than 1% error at a load current of 0.5 A per phase. The results demonstrate a viable alternative that balances accuracy, flexibility, and development efficiency.

ABSTRACT. This paper presents a comparison of two sliding-mode controllers and a standard PID controller for the lateral dynamics of a vehicle, including a BLDC motor as the actuator. The usual application for these controllers is path following for autonomous vehicles. This path is defined as a set of waypoints which are computed by means of sensors mounted on the vehicle as cameras, LIDAR and ultrasonic sensors. A comparison of the performance of the three presented control algorithms is developed in terms of position and orientation tracking errors, the magnitude of the control signals generated for the actuator, and its high-frequency components. In addition, the simulation results and general conclusions about the comparison are presented in terms of commonly used performance metrics.

ABSTRACT. This paper presents a novel control architecture for AC microgrids that enhances the performance of Grid-Forming (GFM) and Grid-Feeding (GFD) converters through coordinated operation. The proposed strategy enables the GFD to operate based on control signals derived from the GFM, allowing for seamless integration of battery charging and voltage regulation. A cascaded PI control scheme is employed to ensure accurate current and voltage tracking, while a hierarchical controller in the GFD manages constant current–constant voltage (CC–CV) battery charging. Simulation results demonstrate that the GFM maintains strict voltage regulation and reactive power support under islanded conditions while simultaneously charging its battery bank, extending its autonomous operation. This approach enhances operational resilience and enables self-sufficient microgrids with reduced dependence on external energy sources. The proposed architecture and control strategy were validated through detailed simulations using various load types, confirming their effectiveness in dynamic microgrid environments.

ABSTRACT. Inverter technology is commonly used in several applications including solar energy. Real-time platforms allow to foretell the inverters behavior which is advantageous along the design process. Therefore, in this paper an analysis regarding the key point for an inverter emulation using Typhoon HIL402 is provided. Conventional topologies as the half-bridge, fullbridge and a multilevel transformerless switched capacitor are considered in this analysis. In the same way, conventional sinusoidal based modulation strategies are used to validate each topology. Finally, results of waveforms for gate signals, output current, output voltage and capacitor voltages are obtained to validate the emulation models.

ABSTRACT. This paper presents the modeling and nonlinear control of a Buck converter using Lyapunov functions. The control strategy is designed in current mode, following a cascade control structure where the inductor current is regulated to indirectly control the output voltage. A Lyapunov-based approach is employed to ensure global stability of the system, regardless of changes in load or reference conditions. The inner loop stabilizes the inductor current using a proportional action derived from a Lyapunov candidate function, while the outer voltage loop includes an integral term to eliminate steady-state error. The resulting control law ensures a fast dynamic response and robust regulation. Numerical simulations in MATLAB/Simulink validate the effectiveness of the proposed method, demonstrating precise tracking of the voltage reference and excellent performance under disturbances and parameter variations.

ABSTRACT. This paper presents the analysis of an interleaved step-up converter topology with Partial Power Processing, applied in direct current microgrids fed by photovoltaic systems. The proposed architecture offers key advantages such as high energy efficiency, better performance at low input voltages and higher power density. The converter performance is described and an average model is developed to facilitate dynamic analysis. To evaluate the behavior of the partial power processing, the Volt-Amperes (VA) Area modeling technique is applied, showing significant benefits compared to a conventional boost converter. Experimental results are presented to validate the theoretical analysis and confirm the feasibility and effectiveness of the proposed solution in DC microgrid environments.

ABSTRACT. This paper presents the modeling and analysis of the Quadratic Bidirectional DC-DC converter which was developed under concept of Redundant Reduced Power Processing (R2P2). The resulting converter shows a quadratic ratio conversion on operation modes, such as Boost Mode and Buck Mode. Based on the analysis of the converter's operation, key aspects such as theoretical efficiency, voltage conversion ratios, and component selection guidelines are defined. Furthermore, simulation results are presented to validate the performance of the proposed topology.

ABSTRACT. Recent technological advances have enabled the collection of digital biomarkers that were previously difficult to obtain and record. Widely accessible devices such as smartphones and smartwatches now facilitate the continuous acquisition of physiological and activity-related metrics, which can be used as digital biomarkers in clinical studies. These biomarkers enable the exploration of correlations and the development of predictive models for identifying potential patterns associated with health conditions. Cognitive impairment, closely linked to physical inactivity, is one of the conditions being investigated. This paper presents the first stage of a broader study aimed at examining this relationship. The initial stage includes clinical evaluations using the International Physical Activity Questionnaire, along with biomarker data collected during six specific physical activities using a smartwatch.

ABSTRACT. The integration of Machine-to-Machine (M2M) communications with fifth-generation (5G) networks has enabled a new era of intelligent, low-latency, and high-reliability applications across diverse domains, particularly in connected healthcare and autonomous medical systems. This work presents a Support Vector Regression (SVR)-based approach for estimating the Angle of Arrival (AoA) in 5G M2M environments, aiming to enhance spatial awareness, dynamic beamforming, and interference suppression in dense deployments. The proposed model is validated using synthetic array data and benchmarked against traditional regression techniques. Results show that the SVR method significantly reduces angular estimation errors under low-noise conditions, demonstrating its robustness and suitability for real-time localization tasks. Moreover, its relevance is underscored through recent clinical use cases of 5G-powered medical platforms—such as capsule endoscopy and telesurgery—where accurate beam control and signal direction estimation are critical. These findings position SVR-based AoA estimation as a key enabler for next-generation M2M medical systems operating in latency-sensitive and spatially constrained environments.

ABSTRACT. This work proposes a convolutional neural network (CNN) implemented in PyTorch for precise angle of arrival (AoA) estimation using uniform linear arrays (ULA). The architecture processes raw complex-valued antenna array data directly, streamlining the signal processing chain and enabling seamless integration into intelligent beamforming systems for high-frequency and millimeter-wave communications. The model was trained over 118 epochs with early stopping, monitoring training and validation losses to ensure convergence and prevent overfitting. Evaluated on separate training, validation, and test datasets, the estimator demonstrated strong generalization, achieving a mean absolute error (MAE) of approximately 0.60° and a mean squared error (MSE) near 1.01 deg$^2$ on the test set. These metrics reflect an average angular deviation below one degree, indicating high precision competitive for antenna signal processing applications.

Representative test examples exhibited pointwise absolute errors under 0.5°, confirming the model’s fine angular resolution. A slight underestimation bias was observed but remained within acceptable limits without significant error spikes, even at the edges of the angular range. A case inference at 25° showed a predicted angle of 25.53°, with an absolute error of 0.53°, demonstrating practical prediction accuracy. Corresponding beamformer phase weights derived from the CNN output validate its applicability for real-time beamforming.

Overall, the combination of low error metrics, stable training convergence, and consistent predictions confirms the CNN’s effectiveness for direct AoA regression from raw antenna data, making it suitable for deployment in real-time high-frequency communication and sensing systems where precise directional information is critical.

ABSTRACT. This study examines the impact of preprocessing techniques, including stratification, oversampling (SMOTE), undersampling, and PCA, on neural network performance for early detection of type 2 diabetes. Using a real-world clinical dataset, multiple preprocessing pipelines were evaluated for accuracy, precision, recall, and F1-score. The optimal model combined oversampling (SMOTE) with a five-hidden-layer neural network, achieving 79.22\% accuracy, 67.19\% precision, 79.63\% recall, and a 72.88\% F1-score. This approach demonstrated the best balance between sensitivity and specificity among tested methods. Although perfect classification was not achieved, results highlight the significant performance enhancement provided by strategic preprocessing in handling limited and imbalanced clinical data. This work offers practical guidelines for developing effective diagnostic models in resource-limited medical contexts.

ABSTRACT. Infrared thermography has proven to be a promising non-invasive tool for evaluating the ocular surface, enabling the analysis of thermal patterns associated with different pathological conditions. This work proposes a logistic regression model for classifying healthy and pathological eyes, using previously selected thermal biomarkers. Thermal images were acquired from 39 patients with various ocular diseases, and three regions of interest were manually segmented: iris-pupil, sclera, and caruncle. Thermal values were extracted from each region at different time points, generating a set of candidate biomarkers. After a selection process based on the individual performance of each biomarker, the three most discriminative ones were identified. These biomarkers were normalized and used as input variables for a logistic regression model. The model was trained using gradient descent and cross-validation, and its performance was evaluated using ROC curves, confusion matrix, and classification metrics. The best model achieved an area under the curve (AUC) of 0.93, with perfect specificity (100\%) and a sensitivity of 80\%, demonstrating strong potential for the automated detection of ocular pathologies. Additionally, class balancing techniques, such as oversampling and undersampling, were evaluated, but no significant improvements were observed compared to the original model. These results suggest that the integration of thermal biomarkers and explainable models such as logistic regression offers a feasible and easy-to-implement alternative for the initial screening of ocular diseases, opening the door for future applications in telemedicine and remote monitoring.

ABSTRACT. This document shows the design and simulation of an exoskeleton prototype to support therapies performed by physiotherapists in the treatment of patients with infantile cerebral palsy (spastic diplegia). The exoskeleton is made using computer-aided design in SolidWorks, based on the extraction of lower limb motion using video analysis techniques with Kinovea software and on the anatomy of 10-year-old children. Human gait is simulated using Matlab's Simulink to determine the velocity, torque, and position parameters that govern the movement of the actuators of the waist, knee, and ankle joints. The results obtained indicate that this video analysis and simulation technique facilitates the selection, design, manufacturing, and programming of the exoskeleton actuators to adjust the kinematic and dynamic analysis to the particular needs and anatomy of children with spastic diplegia.

ABSTRACT. This work validates a modular analog front-end for conditioning surface electromyographic (sEMG) signals in the 20 to 500 Hz band as a building block for actuator control in bionic prostheses. The chain includes a fixed-gain AD623 preamplifier (gain 11), a first-order high-pass filter at 20 Hz, a second-order Butterworth low-pass filter at 500 Hz, and an LM358-based variable-gain amplifier (gain 1 to 11). Each stage was built on a test board and driven with a 0.10 V peak-to-peak sine wave; experimental responses were compared with Multisim simulations. The preamplifier provided 11.6 V/V, about 0.46 dB above nominal. The high-pass filter reached -3 dB at 20 Hz with a 0.4 dB offset, while the low-pass filter measured –1.7 dB at 500 Hz with a maximum deviation of 5.7 dB at 1 kHz. The variable-gain amplifier delivered 1.08 V/V (+0.66 dB) at unity gain and 10.2 V/V (-0.66 dB) at maximum gain, keeping gain error below 1 dB across its range. All gain deviations remained under 6 dB, confirming the design is suitable for integration on a four-layer printed-circuit board with an STM32 analog-to-digital converter for future real-time actuator control in bionic prostheses.

ABSTRACT. Accurate tracking of hands in both human and non-human species is fundamental in fields such as biomechanics and neurology. Artificial intelligence (AI) has emerged as a key tool for achieving markerless tracking with accuracy comparable to manual annotation. In this study, we employed a deep neural network to precisely track user-defined hand movements. By utilizing two synchronized cameras, we obtained data from different angles, allowing for more detailed and robust analysis.

ABSTRACT. Electrical impedance tomography is a non-invasive measurement technique that allows the generation of a graphic reconstruction of a section of a body or surface and from there visualize its impedance profile graphically, in the medical area it has different applications such as the visualization of cancerous tissue in the breast, various strategies have been proposed not only to improve its greatest limitation, its low spatial resolution, but also to be able to visualize the location of the carcinoma in different layers or levels. As a contribution to the development of image reconstruction systems with a greater scope, the use of a three-dimensional model generated from two two-dimensional layers, built by linear interpolation, is proposed. By using this model, it is possible to visualize not only the position of the object, but also the depth at which it is located.

ABSTRACT. This study presents the simulation-based design of a closed-loop thermal regulation system for localized superficial thermal therapy. A simplified First-Order Plus Time Delay (FOPTD) model derived from the bioheat equation is used to emulate tissue response. A PI controller is designed using root locus techniques and evaluated under a family of physiological conditions. The simulation results demonstrate the compliance with clinical requirements regarding rise time, thermal overshoot, and safety thresholds, validating the proposed approach for future experimental implementation.

ABSTRACT. In this paper, an efficient filtering process is proposed to improve the location of oscillation sources in power systems. The proposed method is based on the multiresolution discrete-time Taylor–Fourier transform (DTTFT) designed with O-splines. This multiresolution approach employs a specific design of bandpass filters for the precise extraction of oscillatory components from measured synchrophasor signals. This allows the integration of the Taylor-Fourier filter-bank with O-splines into the well-known dissipating energy flow (DEF) method for oscillation source location, which conventionally uses the Butterworth bandpass filter. The proposed O-splines filter, combined with the DEF method, has been validated considering different scenarios of forced oscillations, where four test cases were created from the well-known Kundur two-area power system to validate the proposal. The attained results confirm that the proposal's advantages compared with the Butterworth filter, since it exhibits an outstanding oscillation source location performance, especially at the beginning of the disturbance.

ABSTRACT. This paper introduces the development of detailed and average models for a battery-based energy storage system (ESS) tailored for electromagnetic transient (EMT) studies using ATP-EMTP. The models, designed with TACS and MODEL elements, include a non-isolated bidirectional DC-DC boost converter and a three-phase, two-level voltage source converter, each represented in detailed and average forms. Simulation results highlight the strengths of each model: the detailed model excels in accurately reproducing high-frequency switching phenomena, while the average model provides computational efficiency by focusing on low-frequency dynamics. These findings demonstrate the trade-off between precision and speed, offering insights for applications ranging from detailed analysis to large-scale grid simulations.

ABSTRACT. In this paper, a state-space approach enhanced by a genetic algorithm is used to design multiple power system stabilizers (PSSs) for damping electromechanical modes. For that, small signal stability is obtained to get a state-space representation, with which a state-space object is formed to obtain the ideal phase lead. Thus, the PSS design consists in match the ideal phase with the state-space object transfer function of the power system stabilizer by means of a genetic algorithm to optimize the parameters T1, T2, T3, T4, and Tw of the Power System Stabilizer (PSS). The primary goal is to align the PSS function with the negative ideal phase behavior of the generators, aiming to place the system poles at a point ensuring optimal damping. This approach seeks to enhance the stability and performance of the power system by achieving effective synchronization between the PSS and generators, thereby optimizing the dynamic response of the system. The obtained results demonstrate the effectiveness of the proposed strategy in improving the dynamic response and stability of generators within an electric power system environment.

ABSTRACT. La integración masiva de energías renovables ha reducido drásticamente la inercia mecánica en los sistemas eléctricos, comprometiendo la estabilidad de frecuencia. Frente a este desafío, los \textbf{inversores Grid-Forming (GFM)} y los \textbf{Sistemas de Almacenamiento en Baterías (BESS)} emergen como tecnologías clave para; los primeros emulan la respuesta inercial de generadores síncronos, mientras los segundos proporcionan ajustes rápidos de potencia activa para mantener la frecuencia en sus valores nominales. En este sentido, este trabajo muestra una revison de estas tecnologias enfocandose en las estrategias de control para GFM y BESS, centrándose en esquemas de inyección de potencia activa y regulación primaria de frecuencia. Se revisan los enfoques propuestos en la literatura para mitigar desafíos operativos, como la coordinación dinámica entre ambas tecnologías y la supresión de armónicos en redes débiles.

Además, se discuten brechas tecnológicas relevantes en áreas como la estandarización, la resiliencia ante fallas, la ciberseguridad, entre otras, identificando oportunidades clave para futuras investigacione. Mediante un análisis comparativo de estudios recientes, este trabajo sintetiza las lecciones aprendidas y resalta las oportunidades clave para futuras investigaciones, enfocándose en cómo facilitar la integración de GFM y BESS en redes con alta penetración de energías renovables.