Offline generation of Selective Harmonic Control PWM (SHC-PWM) solutions for Active Power Filters (APFs) presents a significant computational burden, particularly for extensive operating ranges. This paper proposes an efficient hybrid algorithm combining Differential Evolution (DE) and Nelder-Mead (NM) within a hierarchical framework to drastically accelerate this offline computation. The proposed method minimizes computationally intensive global optimization by adaptively employing fast local search, initialized with previously computed solutions, achieving smooth parameter variations. Achieving speedups greater than 100x compared to other hybrid methods, the algorithm successfully generated millions of unique SHC-PWM solutions averaging 0.07 seconds per point using a standard Intel Core Laptop. APF simulations using the generated data validated effective compensation for controlled current harmonics greater than 99.5%. This work enables the practical generation of large datasets required for advanced SHC-PWM strategies.

Modular Multilevel Cascade Converter (MMCC) have gained wide-spread acceptance in high voltage converter applications. As they require a high amount of capacitors whose voltages must be known for the control of the MMCC, many voltage sensors are needed. This paper proposes the application and the efficient implementation of fixed-point recurrent neural network (fpRNN) based observers for cell voltage estimation of chopper and bridge cells. The proposed observers are capable of learning from data to adjust the estimation model despite its intrinsic nonlinear behavior and unmodeled parasitic effects. The performance of the observer is validated against high-fidelity PLECS simulations. Moreover, the achievable estimation accuracy of different fpRNN observer implementations is compared depending on the utilized fixed-point number format.

This paper presents an adaptive selective harmonic elimination based on artificial neural networks (ASHE-ANN) with indirect model-based online training. The strategy builds on the established analytical formulation of Selective Harmonic Elimination (SHE) to derive a relationship between the magnitudes and phase angles of the Fourier Transform (FT) spectrum and the switching angles, along with their respective error gradients. This relationship enables online training of the artificial neural network (ANN) to dynamically adjust switching angles in real time, achieving precise control of the fundamental voltage magnitude while eliminating target harmonics. Compared to conventional SHE implementations that rely on precomputed lookup tables (LUTs), the proposed approach demonstrates enhanced harmonic suppression under nonideal DC-link conditions (e.g., voltage fluctuations, unbalanced sources). The results highlight the adaptability of the ASHE-ANN strategy in dynamic scenarios, offering an alternative to static LUT-based methods.

A novel and generic frequency-adaptive pole placement (FAPP) algorithm for parallelized proportional, integral and resonant (PIR) control systems is proposed. It allows to place arbitrarily many closed-loop poles at desired locations in the complex plane (usually negative complex half-plane) with individual damping and settling time characteristics. The closed-loop system with parallelized PIR controllers is then capable of tracking any linear combination of constant and sinusoidal reference signals asymptotically. The proposed FAPP algorithm achieves a better damped closed-loop system behaviour than Matlab’s place command. Moreover, even when place fails to place the poles at the desired locations in the complex plane, the proposed FAPP will do the job. Additionally, the PIR controller is equipped with a simple anti-windup strategy and an effective reference feedforward control. The proposed approach is illustrated and validated by several simulation examples including multi-harmonic current control of a nonlinear electrical drive.

Parallel active power filters (PAPF) compensate for current harmonics and grid imbalances caused by (active or passive) nonlinear loads or sources. The proposed novel approach for controlling PAPFs is based on a combination of (i) frequency-adaptive observers for grid voltages and (load) currents, (ii) computation of the PAPF reference currents based on these observed quantities and (pre-)specified filter goals, and (iii) multi-harmonic current control tuned by frequency-adaptive pole placement. The proposed approach achieves grid current harmonic mitigation, load current balancing, and power factor correction. Observers and controllers may adapt themselves to the fundamental grid frequency. High-fidelity simulations of a real-world application show the effectiveness and achievable filtering performance of this overall observer-based control system.

A novel and modular approach for transient response shaping (TRS) of a broad class of nonlinear input/output linearizable systems controlled by nonlinear PI controllers with anti-windup is introduced. It allows to shape the transient response of the closed-loop system to match a prescribed linear reference system (e.g., first-, second-, or nth-order lag system). The derived transient response shaping networks (TRSNs) are simple to tune and guarantee prescribed overshoot, rise time, and settling time. Besides, the TRSNs can be easily and seamlessly integrated into existing PI-based control structures without the need of retuning of the PI controllers, emphasizing the modularity of the proposed approach. Several high-fidelity simulation results, including, e.g., high-performance current control of a nonlinear interior permanent-magnet synchronous machine drive, demonstrate effectiveness, functionality and benefits of the proposed TRS approach compared to classical approaches.

Modular Multilevel Cascade Converters (MMCC) have become a well-established power electronics topology during the past decades since their invention. Among the class of MMCCs, the Modular Multilevel Converter (MMC) is widely adopted nowadays. Due to its modularity, i.e., scalability and flexibility, it is often the preferred choice among other converter topologies. However, the modularity comes with the price of many components (e.g., semiconductors, capacitors, and sensors). The controller must measure and keep every voltage within a specific range to ensure system functionality. This paper proposes a hybrid observer (HO) to estimate all system voltages needed to properly control the MMC. Only current sensors are used, and, in contrast to other methods, even the unmeasured (unknown) DC-link voltage can be estimated. The approach is theoretically derived, and its functionality is validated by high-fidelity simulation under different scenarios. The results prove the excellent estimation performance of the proposed HO.

This paper presents a comprehensive modeling approach for electrical machines with multi-phase systems, incorporating fundamental physical principles and accounting for saturation and cross-coupling effects. It begins with the model’s assumptions and derives a multi-phase winding representation in its orthogonal reference frame. The Clarke factor for multi-phase systems is derived, facilitating the calculation of cross-coupling factors of mutual inductances, especially between different winding systems. The model is further extended to multi-phase windings in an arbitrarily oriented (d, q)-reference frame. The electrical subsystem, which includes multiple multi-phase winding systems, is developed alongside machine power and torque equations. The model is validated through application to a real electrically excited synchronous machine (EESM) with a damper winding, modeled as a multi-phase system. Machine parameters, including nonlinear flux-linkage maps, are obtained through finite element analysis (FEA) to validate the cross-coupling factors of mutual inductances in the (d, q)-reference frame. Additionally, dynamic simulations in MATLAB/Simulink demonstrate the model’s effectiveness in nonlinear current control and dynamic simulation.

This study introduces an optimal reference voltage saturation (ORVS) strategy for nonlinear current control in electrical drives consisting of voltage source converters (VSCs) and electrically excited synchronous machines (EESMs). The ORVS is formulated as an analytically solvable optimization problem designed to maximize feasible current changes while considering both stator and exciter voltage constraints imposed by the limited dc-link voltages of the VSCs. The goal of ORVS is to minimize the difference between the desired and feasible current dynamics’ amplitude while maintaining their direction. The effectiveness of the proposed ORVS is illustrated and validated by high-fidelity simulations of a real nonlinear EESM, demonstrating its advantages over classical reference voltage saturation (CRVS) in achieving better current tracking accuracy and notably reduced cross-coupling effects.

This paper presents a free-wheeling offline and online identification method which allows to parameterize angle-dependent permanent magnet flux linkage prototype functions. The general machine equations, including partial flux linkage derivatives, are formulated, and an analytical integration of the permanent magnet flux linkages is introduced. Furthermore, the harmonics of the measurement signals are identified offline by an axes adjusted discrete fourier transformation (AADFT) and on-line by a frequency-adaptive-observer (FAO). The parametrized trigonometric functions are analytically integrated and transformed from the (a, b, c) to the (d, q, γ)-reference frame. Finally, the proposed identification approaches are experimentally validated by measurements of an interior permanent magnet synchronous machine.

An accurate torque-slip curve estimation is crucial for optimizing the performance of squirrel-cage induction machines, particularly during the design phase of new prototypes. To achieve the best feasible accuracy, it is necessary to capture the operating point dependent nonlinearities of the induction machine as accurately as possible. For this purpose, a novel method is presented which calculates the nonlinear torque-slip curve of squirrel-cage induction machines in an arbitrary (d, q)- reference frame. Unlike conventional approaches which (mostly) assume constant machine parameters, the presented method accounts for magnetic nonlinearities by introducing dense and operating point dependent inductance matrices. The proposed algorithm combines time-harmonic and stationary finite element simulations to estimate torque-slip characteristics efficiently while considering variations in the stator and rotor inductances and resistance. The inductance matrices are obtained from nonlinear stationary simulations, ensuring improved accuracy compared to traditional T-equivalent circuit models. A test model is used to validate the proposed method, showing significant deviations from classical approaches at higher slip values. The results indicate that the nonlinear model provides higher torque estimation accuracy, especially in high slip and breakdown torque regions.

This paper presents a constrained optimization-based neuro-adaptive control (CONAC) for nonlinear synchronous machines (SMs) under voltage constraints, which allows controlling the completely unknown electrical drive system, after a brief learning phase with very satisfactory control performance. The artificial neural network (ANN) in the proposed neuro-adaptive controller (NAC) learns online and empowers the controller to handle parameter uncertainties. Moreover, it solves a constrained optimization problem which allows considering the nonlinear voltage constraints as well, by deriving the adaptation laws of the ANN’s weights from the Lagrangian function. Boundedness of tracking error, convergence of the ANN weights, and satisfaction of the constraints are guaranteed by Lyapunov theory. Numerical simulations in combination with a realistic (nonlinear) synchronous machine drive demonstrate the effectiveness and robustness against parameter and modeling uncertainties of the proposed NAC and its very acceptable constraints handling.

As a crucial characteristic of the grid, its varying impedance affects the performance of grid-connected inverters and, even more severely, system stability. Emulating varying grid impedances to validate inverter control algorithms and system stability offers greater flexibility than physical setups. Conventional power-hardware-in-the-loop (PHIL) systems incorporating grid impedance emulation are confronted inevitably by stability issues and time delays, which require complex modifications. In this paper, a novel PHIL system for grid (impedance) emulation is proposed, where communication between the simulator and controller is established by derivative-free system dynamics. Moreover, an effective method is proposed that eliminates the delay effects and, therefore, also obviates stability issues (almost) completely. High-fidelity simulations validate the feasibility and effectiveness of the proposed PHIL system and highlight the achievable (high) accuracy of the emulation in both transient and steady-state conditions.

Model predictive control (MPC) has emerged as a promising approach for the control of systems with complex, non-linear dynamics, such as modular multilevel converters (MMCs). However, despite the intrinsic nonlinear nature of MMCs, linear MPC is predominantly applied, often without a thorough analysis of how the nonlinearities of the MMC affect control performance and constraint compliance. This paper identifies the impact of the MMC nonlinearities on the control performance of linear MPC and proposes a cascaded approach to mitigate their effects. The proposed approach ensures that the system constraints are still met when linear MPC is applied to the nonlinear MMC. The performance of the cascaded MPC strategy is validated by high-fidelity MATLAB/Simulink simulations and compared to state-of-the-art MPC implementations.

This paper addresses the current saturation challenge in modular multilevel cascade converters (MMCCs). The operation of the MMCC requires the control of ac, dc and circulating currents while accounting for limitations introduced by active or passive components such as semiconductor devices, capacitors, or inductors. Traditionally, current saturation methods focus primarily on limiting the ac currents, often neglecting the effect of their internal dynamics on the cluster currents. As a result, not all current limits are respected. This paper presents saturation methods that provide insight into the cross influences of the three current types, enabling simultaneous enforcement of all current limits. The proposed approaches ensure that the actual currents remain within the prescribed limits during all operating conditions. The effectiveness of the saturation method is validated through high-fidelity MATLAB/Simulink simulations of a high-voltage dc (HVDC) system based on MMCCs.

This paper proposes grid-forming control for doubly-fed induction machines (DFIMs) based on a virtual synchronous machine (VSM), which synchronizes with the grid while providing the desired power response to grid phase jumps. A rotor current control (RCC) based on input-output (I/O) linearization protects the rotor-side inverter (RSI). In addition, a novel transient stator current control (TSCC) reduces current and torque oscillations. The results demonstrate the improved performance of the proposed control compared with (i) no TSCC or (ii) grid-following control using a phase-locked loop (PLL) instead of a VSM for grid synchronization.

In the paradigm of more-intelligent ac drives, the adoption of a system parameter-free algorithm for the current control is a substantial innovation, as it makes modelling and identification of the physical system obsolete. The prediction of current space vector trajectory is based on an adaptive model instead of a traditional motor model. A newly tailored optimisation problem exploits the current prediction to generate the optimal voltage vector reference within an hexagonally limited continuous set. The paper provides several hints for design and implementation, while also providing a remedy for stagnation phenomena that can appear due to the application of a constant voltage during transients. The proposed current control algorithm has been tested extensively on an experimental test rig, featuring a full bus voltage, and two different synchronous motors, either with or without permanent magnets. It is not necessary to fine-tune any controller or to know in advance the motor connected.

A finite-set predictive control (FSPC) approach is designed by considering the discrete-time model of the system under control. Therefore, variations of the model parameters and disturbances due to un-modeled dynamics deteriorate the performance of the FSPC. In order to overcome this problem, this paper proposes a FSPC strategy with an equivalent input disturbance (EID) observer for permanent-magnet synchronous generators (PMSGs) in wind turbine applications. The proposed EID observer estimates the total disturbance due to variations of the model parameters and un-modeled dynamics and considers it in the design of the controller. Experimental results are given to validate the performance of the proposed FSPC scheme.

Efficiency of an IPMSM motor is influenced by the operating point (OP) of the machine. The optimal operating point can be found by using either direct search methods or model-based methods. Model-based methods are sensitive to parameter uncertainties of the equivalent circuit of the drive. In this paper, the impact of various parameter estimation methods on the motor efficiency is compared. Studied methods are Recursive Least Squares (RLS), frequency domain identification at standstill, and the flux linkage map method. Results are compared to direct search methods, where the efficiency is evaluated using the power consumption measurement and direct measurement of the torque. Comparison is performed on a grid of various setpoints (currents and speed). The RLS is tested in two versions of linearization of the flux: one using only the inductances, the second estimating also their offset. Each method is able to obtain good results for some OPs and bad results for other OPs. Overall, good performance was obtained for direct search, offline identified parameters and flux linkage map and RLS with flux offset. RLS without the flux offset does not yield consistent results which implies significantly lower efficiency.

In this paper, an efficient finite control set model predictive control (FCS-MPC) for single-phase multilevel inverter (MLI) is proposed. The MLI topology under control reduces the number of the required semiconductor switches and produces more number of levels compared to traditional and recent topologies of MLIs. This topology has 49 different switching states, which means 49 predictions of the future current and 49 calculations of the cost function are required for each evaluation of the conventional FCS-MPC. Accordingly, the computational load is heavy. Thus, this paper presents a simplified FCS-MPC to reduce the calculation burden by computing the reference voltage and dividing the switching states of the MLI into two sets. Based on the reference voltage, one set is identified to determine the optimum voltage level. The proposed method does not require current predictions. Moreover, the number of cost function evaluations for each iteration is halved. As a result, the execution time is significantly reduced compared to that of the conventional FCS-MPC. The trade-off between the average switching frequency and the tracking performance is also investigated for the proposed scheme and compared to the conventional FCS-MPC. The effectiveness of the proposed solution in transient and steady state is verified by the simulation results.

This paper presents analytic solutions for the calculation of an optimal modulation scheme including frequency variation for a bidirectional single phase dual active bridge (DAB) dc-dc converter used for charging high-voltage batteries of electric vehicles. The proposed modulation scheme facilitates the minimum conduction loss operation of the DAB converter. The losses can be reduced significantly, especially for light- and medium-load conditions. Analytic expressions are derived for a direct calculation of the optimal control parameters. These expressions are verified with a numerical simulation. Moreover, a significant reduction of the necessary computational time can be achieved for the analytic determination.

This paper presents a proportional-resonant (PR) controller for a quasi-Z-source inverter (qZSI) connected to a load via an intermediate LC filter such that it can be used as uninterruptible power supply (UPS) system. In order to improve the performance of the ac side of the qZSI, a PR controller is designed which compensates for selected low-order harmonics. By doing so, the output voltage of the UPS system can be kept sinusoidal not only for linear loads, but also for non-linear loads. For the dc side of the converter, a classical proportional-integral (PI) controller is used to adjust the capacitor voltage to its reference value. The mathematical models of both sides of the qZSI are derived and controllers are designed. In order to examine the qZSI performance under different kinds of loads, simulations are presented verifying the effectiveness of the proposed control strategy.

This paper presents a pruning algorithm using adaptive pruning interval for system identification with general dynamic neural networks (GDNN). GDNNs are artificial neural networks with internal dynamics. All layers have feedback connections with time delays to the same and to all other layers. The parameters are trained with the Levenberg-Marquardt (LM) optimization algorithm. Therefore the Jacobian matrix is required. The Jacobian is calculated by real time recurrent learning (RTRL). As both LM and OBS need Hessian information, computing time can be saved, if OBS uses the scaled inverse Hessian already calculated for the LM algorithm. This paper discusses the effect of using the scaled Hessian instead of the real Hessian in the OBS pruning approach. In addition to that an adaptive pruning interval is introduced. Due to pruning the structure of the identification model is changed drastically. So the parameter optimization task between the pruning steps becomes more or less complex. To guarantee that the parameter optimization algorithm has enough time to cope with the structural changes in the GDNN-model, it is suggested to adapt the pruning interval during the identification process. The proposed algorithm is verified simulatively for two standard identification examples.

In a recent paper, De Jesús proposed a general framework for describing dynamic neural networks. Gradient and Jacobian calculations were discussed based on backpropagation-through-time (BPTT) algorithm and real-time recurrent learning (RTRL). Some errors in the paper of De Jesús bring difficulties for other researchers who want to implement the algorithms. This comments paper shows the critical parts of the publication and gives errata to facilitate understanding and implementation.