This paper provides a comprehensive review of power electronics converter control and energy management for hydrogen production systems through water electrolysis. Hydrogen production from renewable energy sources is a key pathway toward decarbonizing energy systems and enabling large-scale energy storage. Efficient and dependable operation necessitates addressing the dynamic properties of electrolyzers, the intermittent nature of renewable sources, and the coordination among numerous power electronic interfaces. Unlike earlier studies that addressed these aspects separately, this review systematically connects electrolyzer modeling, converter design, control architectures, and energy management to reveal their critical interdependence. By examining these connections, the analysis reveals critical research gaps in real-time coordination, parameter adaptation, and scalable architectures, outlining pathways toward intelligent and grid-independent hydrogen production systems. This review integrates electrolyzer modeling, power converter control algorithms, AC and DC energy hub architectures, hierarchical control schemes, and energy management systems from classical to advanced methods.

This article presents a high dynamic range (HDR) CMOS image sensor, which is designed using a 0.18 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula>m technology, resulting from integrating two previously studied HDR sensor architectures. In this image sensor, we use the readout circuit from a previous design about an HDR CMOS image sensor employing programmable linear-logarithmic counter for low-light imaging applications, combined with the pixel, which is slightly modified, from a previous work about a pixel for asynchronous HDR acquisition through adaptive tone mapping. The pixel resembles the design of a 3T pixel, but includes two more transistors (five in total) in order to allow for the use of the average photocurrent information from the entire pixel array. It is thus possible to obtain an image sensor with programmable sensitivity and HDR. The proposed pixel holds (considering a 10 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula>m ×10 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula> m photodiode) a fill factor of 65%, a pixel pitch of 16.93 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula>m and a pixel area of 15.12 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula>m ×10.16 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mu$</tex-math></inline-formula> m. By configuring sensor parameters, it is possible to perform tone mapping and to control the image brightness, contrast, detail levels and the number of bits used in the image. As a consequence, the resulting image may include either more or less detail, as defined by application-specific (or user-selected) parameters. The simulated dynamic range of the proposed image sensor is equal to 95.72 dB.

TriChewer, a multi-oral-chamber chewing robot, was developed to replicate human chewing trajectories, serving as a tool for food texture evaluation and biomimetic bolus preparation in food science. Cam profiles were designed for vertical chewing with different lateral excursions to provide small, medium and large shear strains during occlusion. The robot's kinematics were modeled, and real human chewing trajectories, measured using an optical tracking system, were converted into motor reference profiles via a closed-form inverse-kinematics transformation derived from the cam-follower geometry, with chamber-specific phase offsets. To enable head-to-head comparisons under identical actuation and ambient conditions, TriChewer operates three oral chambers in parallel, each reproducing a distinct human-inspired force-shear regime. Chamber-specific occlusal-force profiles, paired with distinct trajectories and food types, were imposed to quantify their effects on texture evolution during chewing. A position controller combining velocity/acceleration feedforward with PID feedback ensured accurate trajectory tracking. Experimental studies on roasted peanuts and breadcrumbs were conducted. Results indicate that occlusal force is the dominant determinant of chewing performance. For roasted peanuts, sufficient force progressively reduced particle size, and trajectories combining lateral shear with vertical compression outperformed pure compression. For breadcrumbs, sufficient force with minimal shear decreased thickness and increased hardness and cohesiveness, whereas excessive shear caused tearing and reduced cohesiveness. This study demonstrates the successful application of human chewing motions in a robotic system to analyze food texture dynamics.

In this work, a 650V monolithic enhancement-mode (e-mode) Gallium nitride (GaN) power integration is proposed for industrial applications, integrating ESD protection, a Schmitt trigger, a pre-driver, and a main driver with a high-voltage e-mode GaN HEMT <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R_DS(on)</i> = 120mΩ. To minimize die size and production costs, a compact layout strategy is implemented, where the driver circuitry is strategically placed underneath the high-voltage GaN source pad. For industrial application, the IC was packaged to a compact 8mm x 8mm DFN package for performance evaluation. Experimental results from double pulse tests (DPT) demonstrate a sub-5ns switching time under 400V/6A conditions. Furthermore, the GaN Power IC was evaluated in a 400W hard-switching boost converter, achieving a peak efficiency of 97.9%.

Virtual power plants (VPPs) integrates distributed generation, storage systems, and load under a common control and market framework. Although the VPP concept brings improved flexibility and new services to the utility, it substantially endangers the grid due to the large cyber-physical attack surface. This paper investigates the cyber-physical security of VPPs in modern power systems. First, the structure of VPPs is outlined, including main components, control hierarchies, communication layers, and ancillary services. Further, the cyber and physical vulnerabilities are mapped across different layers of the VPP. This mapping further leads to a structured taxonomy of attacks common in VPP context. Following the vulnerability analysis, the paper looks into the different defense mechanisms relevant to VPP-coordinated distributed energy systems (DESs), in categorization of prevention, detection and isolation, and resilient control. The comprehensive investigation highlights the pressing need for the development of robust, co-designed security and resilience strategies across all layers of VPPs to safeguard their operations against evolving cyber-physical risk, which remains to be a significant gap in the VPP study.

Effectively utilizing flexible energy resources requires optimizing their operation over time to balance dynamic demand and fluctuating supply from volatile renewable sources. Traditionally, this has been achieved through centralized optimization models, which suffer from scalability limitations, single points of failure, and limited flexibility when applied to decentralized and dynamically changing environments. Distributed models offer a promising alternative, providing enhanced flexibility, robustness, and computational efficiency by enabling parallel processing and reducing coordination delays. Thus, this work presents a methodology for the semi-automated transformation of centralized optimization models into distributed architectures, leveraging containerized multi-agent systems to achieve scalable and efficient optimization across multiple computing units. A case study involving 120 electrolyzers distributed across up to 5 optimization agents, including both homogeneous and heterogeneous configurations, demonstrates that the distributed approach accelerates computation time by a factor up to 27.42 compared to a centralized model while accepting a solution quality deviation of only 2.2 %. The optimization integrates site-wide and real-time optimization, ensuring adaptability to fluctuating renewable energy availability and improving system resilience. This combination enables long-term strategic planning while allowing real-time adjustments to maximize renewable energy utilization. The findings highlight the benefits of distributed optimization in modular energy systems and confirm that containerized multi-agent architectures enhance scalability and computational efficiency, making the approach well-suited for real-world applications in decentralized and modular energy networks.

From military applications to everyday devices, hardware (HW) security is more relevant than ever before. The supply chain of integrated circuits is global and involves multiple actors, which facilitate the implementation of various attacks. Its complexity increases the attack surfaces, violating not only the privacy of the users or even national security but also endangering human life. We review some of the publicly known HW attacks that have occurred and propose an assessment scheme for the attacks and the defense on hardware. Using this scheme, we relate the costs of attacks and defense and provide a structured landscape of HW attacks. To illustrate the utility of our assessment scheme, we apply it to a number of real-world and synthetic research cases. We observe a gap between the research use cases and the real-world attacks and envision that the comprehensive assessment of the attacks will enable the development of more suitable countermeasures. Additionally, we revised the security policies for hardware devices, and we conclude that the complexity and obscurity of the supply chain are key parameters impacting hardware security, providing attack surfaces. Finally, we identify the demystification of the supply chain as the main strategy to mitigate this problem.

Utility-scale battery energy storage systems (BESSs) are currently being used to provide auxiliary services, such as frequency regulation, peak shaving, or grid balancing, among others. Hybrid ac/dc distribution grids where the BESS systems are connected in the dc side and the dc/ac interface is implemented through a grid forming (GF) converter are currently researched. These solutions combine the benefits given by the dc distribution and the possibility to provide emulated inertia and damping to the system through the use of GF control techniques. This article presents a novel tuning method, based on small signal analysis, for the configuration parameters of a dual-side GF controller. It aims to minimize the dynamic performance difference between the dual-side and ideal GF controllers, thus ensuring that the dual-side GF provides the expected support to the grid in terms of inertia, damping and primary response, while simultaneously controlling the dc voltage. This is achieved through the optimum tuning of the supplementary dc voltage regulator embedded in the dual-side GF controller. Real-time estimation of the optimum controller gains by making use of an artificial neural network is proposed. Simulation and experimental results are presented to validate the method.