ABSTRACT. Synchrophasors are synchronized phasor measurements obtained from Phasor Measurement Units (PMUs), which provide real-time information on the magnitude and phase angle of electrical variables in electric power systems. However, the transmission of synchrophasor data over communication networks introduces cybersecurity vulnerabilities if the information is not adequately protected.This paper presents the selection, development, and evaluation of an encryption system designed to protect the synchrophasor data stream and measure the associated processing latency. The proposed architecture employs the ChaCha20 cryptographic algorithm in combination with Poly1305, implemented in Python on Raspberry Pi hardware as a secure communication link. Experimental results demonstrate that the system maintains the processing delay below the update interval specified by the IEEE C37.118.2 standard. The embedded hardware implementation validates the integration of a cryptographic mechanism to protect the transmission of synchrophasor measurements without affecting the timing requirements of the electric power system.

ABSTRACT. This work presents a Takagi--Sugeno representation of a reduced neutron--thermal model of an \(8\,\mathrm{MW}\) circulating-fuel molten salt reactor. Six precursor transport delays and one thermal delay are approximated using first-order Padé models, yielding a 17-state augmented system. The remaining nonlinearities are represented by two premise variables and four local subsystems. Validation against the original nonlinear delayed model shows good Padé accuracy and numerical agreement between the Padé and T--S models within the considered domain.

ABSTRACT. This paper presents a simulation-based assessment of a 1.22 MW/2.44 MWh lithium iron phosphate battery energy storage system (LFP-BESS) operated under Mexico’s GDMTH tariff. The modeled system comprises the battery bank, a two-level bidirectional voltage-source converter, an RL output filter, a dynamic industrial load, and an energy management system (EMS) that schedules charging, discharging, and standby operation. Grid current is regulated in the synchronous dq frame by proportional–integral controllers with cross-coupling compensation. A 96 s discrete-time simulation represents two 24 h operating periods with summer and winter tariff schedules. Under the simulated conditions, the state of charge decreases from 100 % to approximately 28 % in summer and to approximately 12 % in winter. The maximum discharge and charge powers are approximately 0.9 MW and 0.6 MW, respectively. The reported current total harmonic distortion does not exceed the 5 % reference threshold adopted in this study, instantaneous efficiency is 98–99 % during steady-state operation, and roundtrip efficiency is 96–98 % within the modeled system boundaries. The main contribution is an integrated electrical and tariff-driven framework for assessing peak-shaving and load-shifting operation in an industrial Mexican scenario.

ABSTRACT. The increasing integration of distributed energy resources and battery energy storage systems has intensified the need for effective power management strategies in islanded AC microgrids. This paper presents the experimental validation of a simplified current-based power management strategy for the coordinated operation of a grid-forming converter and a grid-feeding converter. The proposed approach maintains the conventional cascaded control structure of the grid-forming converter, whereas the grid-feeding converter employs a single current-control loop based on the electrical relationship at the coupling point. Current-sharing coefficients regulate the power contribution of each converter, enabling controlled bidirectional power exchange without requiring additional communication links or hierarchical control structures. The proposed strategy was experimentally validated using a laboratory-scale islanded AC microgrid under different load and power-sharing conditions. Experimental results demonstrate stable coordinated operation, preservation of voltage regulation with low harmonic distortion, and effective bidirectional power exchange between the battery energy storage systems through battery current reversal. These results validate the proposed strategy as a simple and effective solution for enhancing flexibility and energy management in islanded AC microgrids.

ABSTRACT. This work presents a study and analysis of load sharing among DC-DC power electronic converters that are part of a Direct Current (DC) microgrid, which are connected in parallel to meet increasing load demand. The proposed strategy is based on the implementation of a modified PID controller in the voltage control loop, to which a virtual impedance loop is added to improve its performance when converters are connected in parallel. This control strategy enhances the stability and disturbance rejection capability of the microgrid, while also compensating for the effects caused by differences in electrical parameters and variations in operating conditions, thereby achieving a more uniform distribution of current and power among generation sources.

ABSTRACT. This paper presents a machine learning-based platform for forecasting Local Marginal Prices (LMP) in the Mexican electricity market using historical data published by the National Electricity Control Center (CENACE). The platform integrates a PostgreSQL database containing over 209 million records across 2,583 nodes of the National Electric System (SEN), covering the period 2016--2025, a RESTful API developed with FastAPI, and an interactive web dashboard built with React. A global Long Short-Term Memory (LSTM) network with node embeddings was trained on 8.5 million sequences derived from 168-hour input windows, predicting 24-hour LMP profiles for any node of the SEN. The model achieves a coefficient of determination (R^2) of 0.554 and a mean absolute error of 186.18 $/MWh on a held-out temporal test set, outperforming a historical-average baseline across all evaluated metrics. The platform enables hourly and daily LMP forecasting across all nine control regions of the SEN, providing a practical tool for decision-making in electricity markets. Results demonstrate that the integration of deep learning with a full-stack data platform significantly improves accessibility and analysis of electricity price behavior in Mexico.

ABSTRACT. Feature engineering assisted by large language models (LLMs) raises a practical question: is the mutual information (MI) of a generated feature enough to anticipate its contribution to model performance? This work empirically evaluates the relationship between the relative MI of 240 features generated by an LLM (Claude Opus 4.8) and their incremental macro F1 gain (ΔF1) across 10 tabular classification datasets. The ΔF1 of each feature was measured with stratified 5-fold cross validation, and the correlation between MI and ΔF1 was subsequently analyzed per dataset. No general relationship was found: the mean of the 10 correlations is ≈ 0, with opposite signs across datasets (from +0.61 to −0.65). Instead, the correlation of each dataset depends inversely on the headroom of the baseline model (Spearman = −0.79, p = 0.006). With ample headroom, high MI predicts F1 gains; near saturation, the association reverses. The results suggest that univariate MI is a useful selection criterion only when the classifier has room to improve, in line with the theory of feature selection based on conditional information (relevance and redundancy).

ABSTRACT. This study evaluates the tensile mechanical behavior of Acrylonitrile Butadiene Styrene (ABS) specimens manufac- tured using Fused Deposition Modeling (FDM), focusing on the effect of infill density in a Trihexagonal pattern designed in CAD. Unlike structures predetermined by slicing software, the Trihexagonal topology distributes axial stresses in several direc- tions (0◦, 60◦, 120◦). The specimens were fabricated according to ASTM D638-22 TYPE I, evaluating infill densities of 20%, 40%, 60%, and 80%. Tensile tests performed on a SHIMADZU universal machine showed that a 60% density provides the highest ultimate tensile strength (UTS = 19.88 ± 0.94 MPa) and load capacity (1.81 kN), while maximum strain decreased from 5.45% to 4.11% as internal mass increased. It is concluded that the Trihexagonal pattern with 60% infill optimizes the balance between structural stiffness and material usage in FDM components.

ABSTRACT. The evaluation of solar photovoltaic projects in Mexico presents significant operational inefficiencies due to the manual processing of dispersed data from the National Energy Control Center (CENACE). This research developed a full-stack web application designed to automate technical and economic feasibility analysis by integrating hourly solar irradiance and Local Marginal Prices (PML). The methodology employed a Python-based extract, transform, and load (ETL) engine to extract meteorological data from the NASA POWER API and nodal prices from CENACE, storing them in a PostgreSQL database optimized for time-series data. An interactive interface was implemented in React to calculate estimated energy generation, projected income, and the investment payback period in years. Validation was conducted using a paired design with ten engineers, who completed the same tasks manually and with the platform. Three variables were evaluated: processing time, calculation consistency, and analytical precision. Three Wilcoxon signed-rank tests showed significant differences between both procedures. For a 30-day analysis, the manual procedure had a median processing time of 1~h, whereas the platform required 2~s for irradiance and 1~s for PML. Calculation consistency increased from a median of 90\% manually to 100\% with the platform. The median installation cost was $360\,000.00$~USD manually and $382\,457.14$~USD with the platform. The study concluded that the platform effectively centralized strategic information, reduced interpretation errors, and provided a systematic quantitative basis for decision-making in the Mexican Wholesale Electricity Market.

ABSTRACT. Wind-speed forecasting methods often lose accuracy as the forecast horizon increases and generally fail to address target-sensor unavailability and predictive uncertainty simultaneously. This limitation hinders monitoring continuity at the height of interest. To address this issue, this study proposes a multihorizon virtual sensor that forecasts wind speed at 80 m using multiheight information. The method reconstructs and cleans records from the La Ventosa II site, generates temporal descriptors from measurements at 40 and 60 m, and trains Extra Trees, XGBoost, and GRU models. It then combines their predictions with persistence and the power law through a horizon-specific weighted ensemble. The evaluation used 2007 records with a 10-min resolution, chronological walk-forward validation, conformal intervals, and continuous-outage and random-missing-data scenarios. The results showed that the ensemble achieved the lowest RMSE at the 10-min, 1-h, 3-h, and 6-h horizons, reducing the error by 5.3 to 15.9% relative to persistence. The ensemble also achieved empirical coverage rates between 92.9 and 94.9% and showed limited degradation during outages of the 80-m sensor lasting up to 24 h. These results demonstrate that combining multi-height information, complementary models, and prediction intervals support reliable forecasting during upper-sensor unavailability.

ABSTRACT. This paper presents an automated quality inspection system based on edge computer vision for detecting machining burr on CVT Drive-Shaft components. The system implements a two-stage deep learning pipeline deployed on a Raspberry Pi~5 with a Hailo-8L neural processing unit (13~TOPS). The first stage employs a YOLOv11n-seg instance segmentation model that isolates the component from the industrial background, while the second stage uses a YOLOv11n classifier operating on local 96x96-pixel patches extracted from the segmented surface. A temporal consensus mechanism based on majority voting (2-of-3 frames) stabilizes the per-part verdict, reducing transient false positives. Both models are quantized to INT8 and compiled to the Hailo Executable Format (HEF). Experimental results demonstrate a segmentation mAP@0.5 of 99.50% with 100% recall, a patch classification top-1 accuracy of 98.55%, and real-time operation at 30.17~FPS. A 3-hour production trial revealed that the human operator inspected only 54.4% of parts due to fatigue-related omissions, while the system maintained 100% coverage and captured one confirmed burr escape. At a total hardware cost of $501~USD---representing an 80--95% reduction compared to commercial systems from Keyence, Cognex, and OMRON---the system provides an accessible, scalable complement to manual inspection in industrial settings.

ABSTRACT. This work focuses on the design and simulation of a sustainable electric cooking solution for marginal communities based on an isolated microgrid. The isolated microgrid, which is the focus in this paper, is simulated in Matlab/Simulink. A low-voltage step-down DC-DC buck converter is employed to create a DC-link responsible for connecting the PV system to the grid. The DC-link is used to supply a single-phase inverter using a bipolar PWM switching scheme. Further, the PV modules are capable of providing up to 1.755 kW under standard test conditions. The generated energy is used to supply electric household appliances such as an electric stove, a blender, a kettle and a small refrigerator. Further, the remaining portion of the energy generated by the PV array is used to charge a Li-ion battery pack. Simulation results are reported to assess the performance of the whole system in steady-state, demonstrating the capability of the PV microgrid of determining the MPPT condition, charging the Li-ion battery pack and providing a single-phase sinusoidal output voltage with low harmonic distortion. In fact, simulation results demonstrated a %THD at the AC output voltage of just 0.6%.

ABSTRACT. Maintenance decisions for electrical equipment often rely on constant failure rates and direct maintenance costs, neglecting component aging, time-varying demand, and the economic impact of service interruptions. This work extends a renewal-process maintenance model for a low-voltage power circuit breaker by incorporating expected energy not supplied, time-dependent interruption costs, maintenance duration, and hourly industrial demand. Corrective and preventive renewal densities are used to compare three finite-horizon policies: run-to-failure operation, component-wise preventive maintenance, and common preventive maintenance applied to the complete circuit breaker. The policies are evaluated through expected energy not supplied, present value, and equivalent annual cost.

For a 20-year horizon, the equivalent annual costs are 65,025 USD/year for the corrective policy, 61,288 USD/year for component-wise preventive maintenance, and 51,319 USD/year for common preventive maintenance. Thus, the common policy reduces the annualized cost by 21.1\% relative to the corrective case and by 16.3\% relative to the component-wise policy, mainly by coordinating interventions within a single planned outage window.

ABSTRACT. This paper presents a mixed-integer linear programming (MILP) model for the optimal sizing and annual hourly operation of a behind-the-meter photovoltaic (PV) system with a battery energy storage system (BESS), applied to an industrial user under the Mexican GDMTH tariff (Gran Demanda en Media Tensión Horaria). The objective function minimizes the annual cost, including energy, capacity, distribution, and fixed charges, as well as a penalty for unserved energy and the annualized investment costs of the PV-BESS system. The formulation considers the power balance, battery operating limits, charging and discharging decisions, the restriction to prevent grid injection, and the tariff equations used to calculate monthly demand charges. The case study uses hourly load and PV generation profiles for one year, together with monthly tariff values. The model selected 0.702 MW of PV capacity, 0.878 MW of battery power, and 2.374 MWh of battery energy capacity. The results show a 56.08% reduction in the annual cost, mainly due to the decrease in the capacity charge through battery discharge during peak tariff periods. The economic evaluation shows a payback period of 5.25 years.