This paper studies the control-to-output transfer function of the asymmetrical half-bridge with capacitive output filter under voltage mode control. Its knowledge contains valuable information for a responsive and stable controller design. The measured control-to-output transfer functions of an experimental setup under different load conditions serve as the reference. They are compared to a transfer function obtained by time-discrete modeling. Hereby, the duty cycle is perturbed with a multifrequency perturbation signal and is set in relation to the response in the output voltage. While a simplified model delivers only accurate results at low and high frequencies, it shows significant deviations at the resonant peak. Therefore, the influence of different parasitic elements is discussed in time and frequency domain. The study reveals that an accurate description of the asymmetrical half-bridge in frequency domain requires the consideration of the ZVS transition at full load and the inclusion of the current dip – caused by the rectifiers' output capacitance – at light load. Taking these into account, the models match very well with the measurement. Furthermore, varying the input voltage and output power within the specification shows the second-order behavior at all operating points.
This paper introduces a novel and efficient current control solution for induction machines (IMs) utilizing deep symbolic regression (DSR). Existing approaches rely on black-box artificial intelligence model that suffers high computational burden, poor interpretability and challenging real time implementation. The data driven control offers a straightforward yet highly efficient method by developing an optimal control model through training and fitting, which results in an analytically dynamic numerical expression that accurately fit the data. Significantly, this leads to a tractable, concise, and more easily explainable optimal controller. Unlike traditional state-of-the-art proportional-integral (PI) current controllers that depend heavily on specific plant models, the proposed data-driven approach is notable for its independence from explicit plant models. The efficacy of the proposed control solution is validated through simulation and experimental tests using a 3.7 kW induction machine. A controller environment with a TMS320f28388 d DSP is implemented to demonstrate a proposed control performance with a conventional control design. In addition, a closed-loop stability analysis is presented to establish asymptotic stability of the proposed controller over the nominal operating range. The test results suggest that integrating deep learning techniques into power energy conversion applications can easily be realized and can be realized as an explicit, verifiable alternative to black-box controllers for induction machines.
In this paper, a modular all-solid-state impedance-matched pulsed power generator (PPG) is proposed, employing a hybrid charging method that combines the advantages of transformer-based and conventional Marx charging techniques. The architecture of the proposed PPG consists of a multiport active-bridge (MAB) charger, cascaded transformers, impedance-matched Marx generators (IMGs), and a coaxial transmission line. The hybrid charging structure enables distributed energy transfer to each module through multiple dual active bridge (DAB) converters. At the same time, the modular configuration allows flexible scaling of the output voltage by increasing the number of IMGs. A detailed analysis of the MAB charger with cascaded transformers was provided, highlighting the impact of effective leakage inductance on power transfer and decoupling between ports. Experimental validation is carried out using a scaled-down prototype integrating the MAB charger and IMGs. The results confirmed the feasibility of the proposed architecture, demonstrating modular operation and independent control of each port. The prototype successfully achieves the desired repetition rate and output voltage levels, while also revealing practical limitations such as insulation constraints and component stress at higher voltage levels.
The increasing penetration of distributed renewable energy resources has accelerated the adoption of grid-forming converters (GFMCs) as effective replacements for synchronous generators in modern power systems. However, the dynamic interactions among parallel GFMCs, particularly in heterogeneous multi-converter systems, remain insufficiently understood. This paper investigates the interaction dynamics and stability of parallel grid-forming photovoltaic (PV) systems, with particular emphasis on PV-operating-point-dependent dc-side dynamics that locally excite dominant low-frequency oscillatory modes. A comprehensive small-signal state-space model that incorporates PV-source dynamics, converter control loops, and network coupling is developed to analyze the effects of PV operating-point transitions, variations in transmission-line impedance, load conditions, and control parameter mismatches. The analysis demonstrates that PV operating-point transitions from the maximum power point to the constant-current region significantly modify the effective dc-side impedance and decrease modal damping. In parallel systems, these locally excited oscillations propagate through the point of common coupling and electrical network, thereby influencing the dynamic response of neighboring converters. To mitigate these operating-point-sensitive oscillatory dynamics, an active compensator is proposed to enhance local damping and reduce the propagation of oscillations through the parallel system. Nonlinear time-domain simulations and OPAL-RT real-time implementation validate both the accuracy of the developed small-signal model and the effectiveness of the proposed active damping strategy under various operating conditions.
High-electron-mobility GaN power transistors (GaN HEMTs) enable significant improvements in power-electronic converters by enhancing efficiency and power density, making them highly attractive for on-board chargers (OBCs) in electric vehicles. This article presents a numerical optimization methodology for the Dual Active Bridge (DAB) stage of a three-phase modular (a three-phase converter composed of three single-phase converter units), bidirectional two-stage OBC rated at 11 kW. Each single-phase converter unit is composed of an Active Front End (AFE) and the aforementioned DAB converter. In this OBC, component selection and key converter design variables are jointly optimized within a shared design space using a multi-objective genetic algorithm (MOGA), enabling exploration of trade-offs between power density and power conversion losses. Equivalent circuit models capture the inter-dependencies among components within the dual-stage three phase modular converter architecture. The optimization exploits the dual-stage topology to maximize charging energy efficiency by leveraging the modular structure and the variable DC-link voltage throughout the charging cycle. The optimization further incorporates an improved charge-based zero-voltage switching (ZVS) analysis that accounts for variable DC-link voltage operation. A simplified numerical method is presented to construct DAB voltage and current waveforms for optimization calculations, balancing computational efficiency with design precision. The optimization yields Pareto front solutions for charging energy efficiency and power density objectives. A downscaled 5.5 kW prototype — representing half of the intended 11 kW three phase charger rating — was built and tested. The two-stage prototype demonstrates a maximum efficiency of 96.5% at partial load, supporting the suitability of the design for full-power implementation.
With data center compute load currents now in the kiloamp regime, a power delivery architecture enabling integration is desired. This paper presents the design of a class of current-sourced hybrid switched-capacitor converters (hereafter referred to as i/SCCs) for data center 48 V power delivery. Integrating intermediate bus and PoL converters into the load package would allow for a simplified power delivery network and free up space for signal routing into the package. Thus, an integrated power stage would better suit the needs for modern high-performance compute loads. However, while previous power delivery solutions have achieved the high power density and high-efficiency conversion needed for compute-load applications, they face limitations in chip-level integration due to reliance on magnetics as the output filter. To overcome this limitation, we propose an i-SCC design architecture with a current-sourced input stage and no magnetics at the output. By keeping the magnetics at the input, the converter design detailed in this paper presents a class of converters capable of a system-on-chip (SoC) or system-in-package (SiP) integration. In this work, the i-SCC architecture is introduced to establish a new avenue for PoL converter design and a converter topology is developed to validate the concept. The proposed design incorporates a switched capacitor stage with partially soft charged capacitors. Therefore, a derivation for the output impedance is presented. Finally, the presented i-SCC concept is further validated with a 48 V to 6 V, 100 W hardware prototype.
Power semiconductor devices such as SiC MOSFETs and Si-IGBTs are primarily designed for high-efficiency switching between fully on and off states, enabling low conduction and switching losses. However, an increasing number of applications require these devices to operate intentionally in specific operating regions; such as, the saturation region for SiC MOSFETs and the active region for Si-IGBTs, for purposes beyond conventional on/off switching. In these regions, the devices simultaneously sustain significant voltage and current, leading to considerable thermal stress and power dissipation. While many studies examine the behavior of individual device types, a comprehensive review addressing both SiC MOSFETs in saturation and Si-IGBTs in the active region is, to date, still lacking. Motivated by this gap, this paper presents, a unified review covering device physics, electro-thermal behavior, modeling approaches, functional use across diverse applications, and experimental validation, including representative case studies. Key challenges, such as reliability implications and performance limitations under these operating conditions, are analyzed. Finally, major research gaps are identified and the necessary future research directions proposed, offering practical guidance for device selection and converter-level design in modern power electronics.
To evaluate the inertia of inverter-based renewable energy sources, this paper proposes a practical method for identifying two key parameters of a black-box single generator: the inertia constant 2H and the damping coefficient D. The method connects a variable-frequency voltage source to the black-box generation unit, injects a linear frequency disturbance, and measures the resulting active power response. Linear fitting is then applied to the active power waveform to extract its slope and intercept. Based on these quantities, 2H and D are solved jointly, enabling fast and straightforward parameter identification. Additionally, theoretical analysis reveals an inherent estimation error in 2H, which increases with the damping coefficient. To mitigate this error, a compensation method is proposed. After compensation, the estimation error remains below 1% under various conditions. The proposed method enables a high-accuracy inertia assessment using a simple laboratory-level setup.
The rapid scaling of artificial intelligence (AI) workloads is driving a fundamental redesign of the data center power-delivery chain, from the utility grid to the processor die. This review paper presents a stage-by-stage technical benchmarking of state-of-the-art power electronics across the four key conversion stages that define this chain: the power supply unit (PSU), the high-voltage intermediate bus converter (HV IBC), the on-board low-voltage intermediate bus converter (LV IBC) and voltage regulator module (VRM), and the wide-bandgap (WBG) semiconductor devices that enable each stage. For the conventional AC-distributed architecture, totem-pole power factor correction (PFC) and LLC resonant converter topologies are benchmarked across the 3–12 kW power range. The analysis identifies the component-scaling limitations that constrain PSU power density and motivate the transition toward high-voltage direct-current (HVDC) distribution. For the emerging 800 V HVDC architecture, input-series-output-parallel (ISOP)-based GaN LLC converters are shown to achieve power densities exceeding 20 times those of conventional PSUs while maintaining comparable efficiency. In addition, fault-protection challenges unique to DC distribution are systematically analyzed. At the board level, LV IBC, VRM, and direct 48 V-to-point-of-load (PoL) conversion architectures are comparatively evaluated in terms of efficiency, power density, transient response, and thermal performance. This evaluation establishes a topology-to-performance mapping and provides a structured comparison between conventional two-stage and emerging single-stage power-delivery architectures. At the device level, commercial 1200 V SiC MOSFETs are benchmarked against 650 V GaN HEMTs for HVDC applications, while low-voltage Si MOSFETs are compared with 15–100 V GaN HEMTs for board-level conversion stages. Key figures of merit, including Ron×Qg, package area, current density, and cost per ampere, are used to quantify where wide-bandgap technologies have surpassed silicon and where challenges related to cost, current capability, and packaging maturity continue to limit adoption. By consolidating circuit-level and device-level benchmarks within a unified grid-to-chip framework, this review identifies the dominant efficiency–power density–voltage–cost tradeoffs across each conversion stage and provides insights into the development of a scalable, high-density power-delivery ecosystem for next-generation AI server racks.
Wireless Power Transfer systems for Electric Vehicle charging are usually designed and controlled with the primary objective of maximizing delivered power and efficiency. However, the same electrical quantities already available in the power converter can also carry information about the physical state of the magnetic connection. This paper investigates the concept of wireless power transfer as a sensor, where inverter-side voltage and current measurements are exploited to infer relevant operating conditions of an electric vehicle wireless charging system without adding dedicated sensing coils or external positioning sensors. A primary-side electrical signature is defined from the fundamental components of the inverter voltage and current, including fundamental voltage and current magnitudes, equivalent-impedance magnitude and real and imaginary components, voltage-current phase displacement, and a fundamental active-power indicator. The approach is experimentally evaluated on an electric vehicle oriented wireless power transfer setup equipped with series-series compensation and a battery-emulating electronic load at 350 V. Four physical states are considered: aligned pads at 125 mm air gap, aligned pads at 175 mm air gap, combined lateral misalignment of 100 and 75 mm, and insertion of a metallic object at the pad center. The results show that the considered operating conditions generate distinguishable inverter-side signatures. Air-gap variation and lateral misalignment mainly affect the equivalent impedance magnitude, current amplitude, and phase displacement, while preserving the DC-side transferred power close to the nominal value. Conversely, the metallic-object condition produces a negative equivalent-impedance phase of $-2.38^\circ$ and a 20.6% reduction of DC-side transferred power. This indicates that metallic-object insertion cannot be interpreted as a simple reduction of magnetic coupling. An exploratory window-based analysis further shows that the proposed primary-side features form separated clusters in the measured dataset. Since the available dataset includes a limited number of acquisitions per condition, the classification is not claimed as statistically exhaustive, but it demonstrates the feasibility of using converter-side electrical signatures as a diagnostic layer for future self-aware EV wireless chargers.
The transition toward converter-dominated distribution grids, driven by distributed generation, storage, and increasingly nonlinear demand, is intensifying power-quality disturbances and exposing limitations of passive mitigation. Unified Power Quality Conditioners (UPQCs) are power-electronics-based compensation devices that integrate series and shunt converters through a common DC-link to simultaneously regulate load voltage and shape source current, but practical deployment is largely determined by robust control and effective fault protection. While the UPQC literature includes many surveys centered on converter topologies, a consolidated review of control and protection remains comparatively limited. This paper fills that gap by organizing control strategies into a layered framework that links reference generation, synchronization and DC-link regulation, converter-level tracking, and modulation, emphasizing representative solution families and their practical trade-offs. Protection is reviewed with a focus on the series converter and DC-link, including bypass and fault-current-limiting concepts, and complemented by discussion of shunt current limiting under stressed conditions. Finally, the paper highlights open challenges and research directions aimed at improving UPQC reliability and accelerating adoption in modern distribution networks.
Monitoring junction temperature is crucial for ensuring the reliable operation of power devices. Although numerous solutions have been proposed over the years, real-time junction temperature sensing remains a challenging task. The recently proposed temperature-sensitive optical parameter junction temperature sensing using emitted electroluminescence offers many advantages, such as galvanic isolation, real-time implementation and is not influenced by the degradation of the package. This work focuses on junction temperature extraction under realistic power converter switching conditions investigating how the switching frequency and current ripple across the semiconducting device may affect the acquired luminescence. Furthermore, to discover the underlying mechanisms and provide an accurate junction temperature detection without neglecting the switching conditions, a polynomial model which allows real-time junction temperature sensing is built and fitted to the experimental data. With the proposed method, we achieve 2 kHz bandwidth temperature sensing with maximum 5°C error at 50 A average current.
Finite-control-set model predictive torque control of multiphase induction machines is often limited by heuristic tuning of weighting factors and inadequate secondary-plane current regulation. To address these limitations, this paper proposes a weighting-factor-free sequential model predictive torque control strategy with virtual voltage vector synthesis (SMPTC-VV) for six-phase induction machine drives supplied by dual two-level voltage source inverters. The proposed scheme employs a sequential optimisation structure that removes weighting factors from the cost function formulation. Virtual voltage vectors are incorporated to suppress non-torque-producing currents in the $x-y$ plane, enabling simultaneous regulation of electromagnetic torque, stator flux, and secondary-plane currents. Experimental validation under steady-state and dynamic conditions, including $\pm$25% magnetising inductance mismatch, demonstrates accurate torque and flux regulation, fast transient response, and improved current quality. The proposed strategy achieves improvements of 10.08% and 53.41% in $\text{THD}_\alpha$, and 54.26% and 69.21% in $\text{THD}_{a}$, compared to SMPTC and classical PTC, respectively. In addition, flux regulation is improved by 28.03% and 31.13% relative to SMPTC and classical PTC, respectively, while non-torque-producing $x-y$ currents are improved by 75.40% and 82.58% with respect to SMPTC and classical PTC. The computational effort is limited to 966 floating-point operations per sampling period, corresponding to improvements of 50.66% and 75.66% relative to SMPTC and classical PTC, respectively. The results demonstrate an efficient and experimentally validated predictive torque control solution for high-performance multiphase drives.
This paper outlines a new solution based on the B4 topology for reducing leakage currents in three-phase, grid-connected photovoltaic inverters. The effectiveness of this topology in reducing leakage currents, which are primarily caused by the relatively high capacitance to ground of the PV modules, is evaluated using both symmetrical and asymmetrical grid connection LCL filters. The results show that using an asymmetric filter is particularly effective in reducing leakage currents. The study is complemented by a presentation of the modelling and control procedure for grid-connected B4 inverters with an asymmetric filter, which differs from the conventional approach. The concepts introduced have been validated through experimentation involving the measurement and comparison of the leakage current produced by the B4 topology with that produced by conventional B6 topologies, as well as other previously proposed solutions. Additionally, the dynamic response and grid current distortion of the converter were measured to demonstrate the high performance of the proposed solution.
This paper presents a high-step-up DC-DC converter suitable for low-voltage source applications, such as renewable energy systems. The proposed topology integrates a two-stage structure, two coupled inductors with dual windings, and a voltage lift technique to achieve high voltage gain, continuous input current, and reduced semiconductor device voltage stress. The converter features a common ground between the input and output sides, which facilitates system integration and control. The input current is divided between the two coupled inductors, thereby limiting losses in the input-side components. Steady-state analysis of the converter was performed to determine the voltage gain, voltage stress, and component currents. Relationships relevant to the converter's design have been presented to facilitate its operation under the required conditions. Additionally, small-signal analysis was conducted to support the controller design, and the corresponding transfer functions were derived. The converter's power losses have also been calculated. Afterward, a comparison is presented between this structure and other designs in the literature to identify its advantages and disadvantages. To validate the theoretical analysis and practical feasibility, a 100-W prototype with a 200-V output voltage was implemented. The experimental results show good agreement with the theoretical predictions.
This paper presents a novel high-boost active-switched quasi-Z-source inverter (qZSI) topology, specifically designed for renewable energy systems and hybrid electric vehicle applications. The proposed inverter achieves a high voltage gain (B) at a low shoot-through duty ratio (DST) while enabling a high modulation index (M). Compared to conventional active-switched qZSI structures with an identical number of passive and active components, the proposed topology provides a significantly improved voltage gain, particularly in the low shoot-through duty ratio region. Key advantages of the proposed design include continuous input current, reduced voltage and current stresses on both active and passive devices, and high conversion efficiency. Furthermore, unlike the conventional Z-source inverter, the proposed configuration eliminates inrush current at startup and exhibits a low input current ripple. Additional benefits include the presence of a common ground between the input source and inverter legs, reduced size of the impedance network inductors and capacitors, and enhanced output voltage quality enabled by the use of a higher modulation index. The paper presents the operating principles, a detailed theoretical analysis, and comparative performance evaluation against conventional counterparts in terms of voltage gain, component count, and voltage/current stresses. The superiority of the proposed topology is demonstrated through analytical results and verified experimentally. Laboratory results confirm the validity of the theoretical analysis and the effectiveness of the proposed inverter under various operating conditions.
Digital modeling plays a critical role in digital twin systems for back-to-back modular multilevel converters (BTB MMCs), enabling optimal design, condition monitoring, and lifetime prediction. This paper proposes a discrete-time digital model of BTB MMC system, including a power-circuit model and a loss-thermal model, to evaluate key performance indices such as voltage and current stresses, power losses, and junction temperatures of power devices. A generic modeling approach is developed to enable the flexible integration of different converter topologies, modulation strategies, and power device selections, allowing efficient evaluation of various MMC configurations. In addition, analytical formulations of system parameters are derived based on an explicit backward Euler method, eliminating iterative calculations and significantly improving computational efficiency. Simulation and experimental results validate the accuracy of the proposed model and demonstrate its superior execution speed compared with commercial simulation platforms and conventional digital models.
Resonant converters, specifically those based on the CLLC topology, are essential for applications requiring galvanic isolation and wide operating ranges, such as bidirectional Electric Vehicle (EV) charging. However, their performance depends heavily on design parameters that often conflict when the converter is optimized for both forward and reverse power flows. This paper addresses this challenge by proposing a new design methodology for bidirectional resonant systems that simultaneously considers both power-flow directions to maximize the $h_{\mathrm{f}}=L_{\mathrm{m}}/L_{\mathrm{r1}}$ ratio, simplifying the integration of the resonant inductances into the transformer and therefore reducing the component count and potentially increasing the power density of the system. A reconfigurable resonant tank architecture (C4LC) that adds an extra inductor ($L_{\mathrm{p}}$) on the primary side of the tank via a switch during reverse operation is presented, allowing the converter to be optimally designed for battery charging (forward mode) without being constrained by reverse mode requirements. Experimental results, based on a 1.3 kW prototype, show that this reconfiguration increases voltage-gain range in reverse operation by an average of 67%. In addition, the methodology yields an $h_{\mathrm{f}}$ value nearly three times larger than in conventional configurations and allows for easier integration of the resonant inductors into the transformer.
This addresses errors in [1]. Due to a production error, the matrix (13) and (14) are not displayed correctly.
In the conventional modulated model predictive control (MMPC) methods, the output voltage vector is synthesized from a fixed candidate vector set, which limits the synthesizable output voltage vectors. This limitation results in a large voltage error between the reference voltage vector and the output voltage vector, which degrades steady-state performance. To address this drawback, this paper presents an enhanced MMPC method for permanent magnet synchronous motor (PMSM) drives using an extended candidate vector set. In the proposed MMPC method, the candidate vector set is constructed using an adaptive voltage vector (AVV), which is geometrically derived from the reference voltage vector. This approach overcomes the limitation imposed by the fixed candidate vector set, thereby reducing the voltage error between the reference voltage vector and the output voltage vector. As a result, the proposed method improves steady-state performance and reduces current ripple over a wide operating range, without a significant increase in computation time. The effectiveness of the proposed method was experimentally validated using a 15 kW PMSM drive system with a 2-level voltage source inverter.
