Nenad Mijatovic

Power converters with a small dc-link capacitor are designed to increase the lifetime by replacing bulky electrolytic capacitors or power factor correction circuits, presenting an attractive solution. However, when these power converters are connected to a weak grid, a resonance can occur between the line-side impedance and the dc-link capacitor. This resonance can lead to issues such as increased total harmonic distortion (THD) of the line current due to ac line harmonics. This article proposes an advanced finite control set model predictive control (FCS-MPC) methodology with a multistep ahead prediction horizon for a three-phase diode rectifier system equipped with a small dc-link capacitor, driving a permanent magnet synchronous motor (PMSM). The FCS-MPC cost functions are designed to control the currents on both the load and dc-link sides, aiming to reduce THD and suppress harmonics in the grid-side current. To adapt the system to varying operational conditions and ensure stability under variable load, torque, or speed conditions, two brain emotional learning-based intelligent controllers (BELBIC) are employed to design the weighting factors, aiming for a significant reduction in harmonics on both the grid and load sides. The effectiveness of the proposed method is validated through experimental tests.

The approximation of model predictive control (MPC) algorithms with neural network models has received significant attention in the power electronics research community, as a valuable tool to enable the real-time implementation of computationally expensive methods. Training machine learning models to imitate MPC algorithms typically follows a traditional supervised learning flow, with a static training data set generated through either uniform sampling or simulation. Training with uniformly distributed data allows models to maintain consistent performance throughout the entire operating space, but very small models may fail to achieve satisfactory results in the regions corresponding to expected operation. Conversely, simulation data can be used in training to obtain models that can accurately track certain trajectories but fail to achieve adequate performance in underrepresented regions. The present article proposes a method to combine the advantages of both methods, by training models on sampled data with importance weights obtained through the kernel density estimation of experimental data. This allows the training process to target specific regions of the operating space, and results in measurable performance improvements. The method is demonstrated through the imitation of a moving discretized control set MPC algorithm for a dual active bridge converter, and is validated through experimental results.

Ensuring the reliability of power electronic converters is a matter of great importance, and data-driven condition monitoring techniques are cementing themselves as an important tool for this purpose. However, translating methods that work well in controlled lab environments to field applications presents significant challenges, notably because of the limited diversity and accuracy of the lab training data. By enabling the use of field data, online machine learning can be a powerful tool to overcome this problem, but it introduces additional challenges in ensuring the stability and predictability of the training processes. This work presents an edge computing method that mitigates these shortcomings with minimal additional memory usage, by employing an autonomous algorithm that prioritizes the storage of training samples with larger prediction errors. The method is demonstrated on the use case of a self-commissioning condition monitoring system, in the form of a thermal anomaly detection scheme for a variable frequency motor drive, where the algorithm self-learned to distinguish normal and anomalous operation with minimal prior knowledge. The obtained results, based on experimental data, show a significant improvement in prediction accuracy and training speed, when compared with equivalent models trained online without the proposed data selection process.

The present paper identifies the root causes of current ringing in a Dual Active Bridge (DAB) power converter and proposes a novel methodology to design the magnetic tank (transformer, primary and secondary inductances) configuration to minimizeit. In particular, two metrics proportional to the current harmonic content on each transformer side are correspondingly proposed based on a steady-state frequency analysis of primary and secondary ac voltages and impedance of the magnetic tank. To get the metrics, different combinations of dv/dt, switching frequency, primary and secondary inductance configuration, and stray capacitance of the magnetic tank (model inputs) are simulated to generate primary and secondary harmonic current indicators (model outputs). Then, a machine-learning model is trained to predict the harmonic current metrics for each given operating setpoint. The model is able to make predictions beyond its training data, therefore generalizing to unseen inputs and enabling rapid evaluation of various designs. Experimental results are presented to evaluate the performance of the metric model for 4 different magnetic tank configurations. Designers can use the methodology to rapidly analyze the generation of current ringing under multiple operating parameters, identifying operating zones with low harmonic content and reduced sensitivity to parameter shifts for the converter's operation

This article proposes a novel attack identification mechanism for Internet of Things-based (IoT-based) converter-composed dc grids, where each agent collects its own and neighbors’ measurement data for output regulation to meet a preceding power-sharing consensus. Independent from model-free or average-model-based attack detection theories, this mechanism is mainly inspired by converter stitching behavior analysis. Correspondingly, when facing latent signal substitution or agent instigation attacks, through comparing estimated signals with received ones for signal source authentication, both self-sensors and neighbors will be inspected. Eventually, not only can such attacks be detected but also will respective attack sources be identified. A simulation case of 4-agent 800-V IoT-based dc grid on Simulink and a hardware case of 3-agent 90-V IoT-based dc grid on the dSpace testing platform were investigated. Experimental results revealed that the estimation ratio error kept lower than 3.9% and all attacks were successfully identified, verifying the effectiveness of the proposed mechanism.

This paper proposes a real-time algorithm for identifying the grid parameters, which is concurrently used for online tuning of the predictive controller in each iteration, in a gridtied active front end (AFE) voltage source converter (VSC) applications. The algorithm is designed by inspiring from the concepts of the extended Kalman filter (EKF) and the model predictive controller (MPC). The performance of the algorithm highly depends on the weighting factors of the algorithm. The artificial neural networks (ANN)-based algorithm is used to find the optimal set of weighting factors among the ones in a parameter search block. An offline particle swarm optimization (PSO) is run to provide the data source for the parameter search block. The algorithm identifies not only the inductance but also the resistance of the grid. Additionally, the hard constraints on the amplitude of the input and output variables are guaranteed. The validation of the proposed approach is performed experimentally and compared with the state-of-the-art conventional methods. The experimental results show the proposed method could effectively stabilize the system in weak grid conditions and under wide impedance variations. Additionally, the accuracy of the proposed impedance identification method is 96%.

This article proposes a distributed intelligent secondary control approach based on brain emotional learning-based intelligent controller (BELBIC) for power electronic-based ac microgrid (MG). The BELBIC controller is able to learn quick-auto and handle model complexity, nonlinearity, and uncertainty of the MG. The proposed controller is fully model-free, indicating that the voltage amplitude and frequency deviations are regulated without previous knowledge of the system model and parameters. This approach ensures low steady-state variations with higher bandwidth and maintains accurate power-sharing of the droop mechanism. Furthermore, primary control is realized with a robust finite control set-model predictive control in the inner level to increase the system frequency bandwidth and a droop control in the outer level to regulate the power-sharing among the distributed generations. Finally, experimental tests obtained from a hardware-in-the-loop testbed validate the proposed control strategy for different cases.

This article proposes a robust cooperative distributed secondary control strategy for dc microgrids, with the main focus on reducing communication burdens. To this end, we adopt a sparsity-promoting consensus-based distributed secondary control framework for converter-interfaced dc microgrids consisting of multiple distributed generation (DG) units. The proposed control strategy relies on less (sparse) information exchange over a communication network and does not require exchanging voltage signals or their estimated values amongst neighboring DG units. Rigorous Lyapunov-based stability certificates for dc microgrids are derived. The stability conditions can be verified without the knowledge of distribution lines connecting different DG units. The effectiveness of the proposed secondary control approach is verified by experimental results and a comparative simulation case study on a multiple-DG dc microgrid.

The nonlinear behavior of constant power loads (CPLs) and their negative incremental impedance, which impose adverse effects on system damping and stability margins, necessitate developing proper strategies for the efficient implementation framework of a dc microgrid. In this article, a fixed-time sliding-mode disturbance observer is first addressed to provide an estimation of the instantaneous power flow moving along the uncertain CPLs with time-varying nature within a fixed time. It not only expedites the estimation rate but also improves the robustness against physical parameter variation. The fast fixed-time backstepping technique is, then, developed based on the estimated load power to control the duty cycle of the boost converter such that the entire power grid becomes stable and the desired voltage of the dc bus is tracked within a fixed time irrespective of the initial conditions. A rigorous Lyapunov-based approach is represented to guarantee the fixed-time stability of the proposed scheme. Finally, to validate the merits and implementation feasibility of the proposed methodology, the experimental realization is represented under various operating case studies.

Sharp LC resonance peak may enlarge inherent harmonics, dominant harmonics of the supply voltage generated by a three-phase diode rectifier, around the resonance frequency, which will intensify fluctuations of the dc-link voltage, and grid currents in motor drives with small dc-link capacitor. In this article, a novel resonance suppression strategy based on inherent harmonics impedance shaping is proposed specially for three-phase input permanent magnet synchronous motor (PMSM) drives with small dc-link capacitor. Instead of increasing the damping of the whole system for resonance suppression, the motor side power disturbance can be reduced by designing the impedance for inherent harmonics around LC resonance frequency based on the new model with the consideration of motor side characteristic. For damping current generation, the angle of the virtual admittance is modified to eliminate the influence of the dc-link voltage sampling delay. Experimental results show the effectiveness of the method on a prototype of electrolytic capacitorless PMSM drive.