Ascensión Zafra-Cabeza

La estimación del estado de salud (SOH) es esencial para la operación de baterías de iones de litio. Aunque las redes LSTM pueden alcanzar una alta precisión, comúnmente utilizan indicadores temporales, por lo que proporcionan una interpretabilidad limitada. Este trabajo propone un sistema de inferencia neuroborroso adaptativo (ANFIS) entrenado con características electrotérmicas de cada ciclo, las cuales son extraídas de las señales de voltaje, corriente y temperatura. Usando el conjunto de datos de degradación de baterías de Oxford, así como una definición común del SOH, se llevó a cabo una comparación entre ANFIS y un modelo LSTM base que incluye el índice de ciclo. Los resultados muestran que ANFIS alcanza una precisión competitiva sin utilizar información explícita de indicadores temporales, además proporciona transparencia a través de las funciones de pertenencia, la activación de reglas y el análisis de sensibilidad, por lo que se presenta como una alternativa práctica y explicable para el monitoreo del SOH de baterías.

This paper presents a fault-tolerant model predictive control (MPC) strategy applied to a microgrid with renewable generation and battery energy storage. The proposed approach combines a zone-based MPC controller for the battery state of charge with a fault detection and diagnosis (FDI) module. The diagnosis compares the measured battery behavior with the behavior computed by a reference model, obtaining a functional estimate of the battery state of health (SOH). This estimate activates a reconfiguration strategy that updates the internal model of the controller and modifies the admissible operating band of the state of charge. The architecture is evaluated through simulation under different battery degradation levels. The results show that the proposed strategy distinguishes between scenarios with and without diagnostic activation, improves the performance of the reconfigured controller compared with the non-reconfigured operation, and reduces the deviations of the state of charge with respect to the admissible operating zone while satisfying the power constraints of the system.

This work presents the upgrade of the FOX hybrid electric vehicle as an experimental platform for embedded control testing, energy management, communications, HMI supervision, and validation of hardware and software architectures. The original platform was based on a distributed PC/104 architecture, with a control unit running on QNX and an independent supervisor running on Windows XP/LabVIEW. The proposed upgrade replaces this configuration with an ATC8110 industrial ECU based on Linux RT. This architecture integrates vehicle signal acquisition, communication with the BMS and motor controllers, selection and execution of control algorithms, command transmission to the actuators, operation through an HMI developed in Python/Qt, and experimental data logging. In addition, it incorporates safe start-up logic, a modular software structure, and continuous integration workflows using GitHub Actions. The resulting platform provides a flexible basis for continuing future tests on control, diagnostics, energy management, and experimental validation in the vehicle.

This work presents a novel two-layer distributed control system for energy management in an energy community, integrating stochastic and distributed model predictive control with a fault-tolerant mechanism. Unlike conventional approaches, the proposed method allows agents to autonomously detect, isolate, and adapt to faults in a fully distributed manner, without relying on a central coordinator. This is achieved through a smart contract deployed on a blockchain network, which facilitates decentralized coordination and fault management. The lower layer implements a fault-tolerant mechanism that detects, isolates, and adapts the control parameters of affected agents. It minimizes false positives through residual signal analysis and dynamically adjusts operational constraints to maintain system stability and performance. When multiple agents are affected, fault-related information is exchanged among them, enabling a coordinated response. This allows proactive actions such as disabling power lines linked to faulty agents or adjusting power transactions in response to reduced output capacity from a neighboring agent. The upper layer employs chance-constrained model predictive control to handle uncertainties in energy demand and renewable generation, ensuring robust decision-making under variability. Energy scheduling is managed through blockchain-based smart contracts, providing a secure and decentralized coordination framework. The effectiveness of the proposed scheme is validated through simulations, demonstrating its ability to optimize energy distribution while enhancing fault resilience in energy communities.

Este artículo presenta una formulación detallada del control predictivo basado en modelo con escenarios ponderados (WS-MPC, por sus siglas en inglés) orientada a sistemas lineales con incertidumbres. El enfoque propuesto combina técnicas de optimización robusta y generación de múltiples escenarios dinámicos del sistema, asignando una ponderación diferenciada al escenario de peor caso dentro del problema de control. Esta estrategia permite una mayor resiliencia ante perturbaciones estocásticas y modelado incierto, mejorando la capacidad del controlador para anticiparse a comportamientos no esperados. El WS-MPC adapta dinámicamente las acciones de control en función de la evolución de todos los escenarios considerados, manteniendo la factibilidad y estabilidad del sistema ante condiciones desfavorables. Los resultados de simulación demuestran que el enfoque propuesto incrementa la robustez operativa del sistema y proporciona una respuesta más segura y confiable frente a variaciones inesperadas. En particular, este controlador se ha implementado en un caso de estudio académico centrado en la gestión del inventario de una tienda.