This paper introduces a novel fault diagnosis approach for two-level Voltage Source Inverters in Motor Drives, using Shallow Neural Networks. A key feature of this study is the implementation of three different Classifiers: two are designed to both detect and diagnose faults, while the third is specialized in fault detection and localization. The strategic adoption of Shallow Neural Networks plays a pivotal role in substantially reducing computational demands while enabling the direct utilization of raw, normalized data obtained from both simulation and experimental activities. The research includes a comparative analysis of three distinct shallow neural network architectures against other machine learning methods to underscore their effectiveness. Specifically, the use of shallow Long Short Term Memories networks is a pioneering effort in the field of power converter fault detection, especially for their reduced memory requirement (one hidden layer with 64 units). The final results show that the average time per observation for the localization classifier(7 Class-Classifier) across the entire test set is equal to 39 $\mu$ s. This efficiency is coupled with promising metrics, including a precision, recall and F1-score of 94.50%, 94.24%, and 94.35%, respectively.

This paper investigates an innovative modulation technique for a predictive control applied to a synchronous reluctance motor drive. The new formulation of the duty cycles, in both linear and overmodulation regions, relies on the identified flux-versus-current characteristics. Furthermore, integral terms have been introduced in the predictive control to achieve satisfactory reference tracking performance with zero steady-state error, even under model parameters mismatches. Low current ripple with smooth and fast dynamic responses are achievable at fixed switching frequency over the whole current operating range. Simulations and experimental evidence show the effectiveness of the proposed controller against standard controllers such as field-oriented current control and deadbeat current control, guaranteeing low current ripple, robustness against parameters variations, and fast dynamic performance.

The scientific literature acknowledges cascaded H-bridge (CHB) converters as a viable alternative to two-level inverters in electric vehicle (EV) powertrain applications. In the context of an electric vehicle engine connected to a DC charger, this study introduces a state of charge (SOC)-governed method for charging li-ion battery modules using a cascaded H-bridge converter. The key strength of this algorithm lies in its ability to achieve balanced charging of battery modules across all three-phase submodules while simultaneously controlling the DC charger, eliminating the need for an additional intermediate converter. Moreover, the algorithm is highly customizable, allowing adaptation to various configurations involving different numbers of submodules per phase. Simulative and experimental results are presented to demonstrate the effectiveness of the proposed charging algorithm, validating its practical application.

This paper presents a new topology for EV powertrain, called Reconfigurable Cascaded Multilevel Converter, able to simultaneously implement power conversion and active battery system management. The latter feature is performed through the Reconfigurable Battery Module structure, where the serial connection of cells is controlled through a pattern of switches. This work contributes to the improvement of EV powertrain multilevel topologies by adding the advantages of reconfigurable systems. Compared to other multilevel topologies, an optimized control algorithm allows maintaining the converter efficiency comparable to more standard solutions and, concurrently, performing sorting algorithms to keep balanced SOC dynamics between the cells. Moreover, the new topology can be completely customized, according to different application requirements. To validate its feasibility in terms of efficiency, the converter is compared to other three alternative topologies. Efficiency measurements are carried out to validate the comparison. Finally, experimental results show how the active management of the battery stack can avoid voltage imbalances and be resilient to sudden voltage or load changes.

With the increasing integration of renewable energy sources into electrical grids, energy storage systems have become crucial for stability and regulation. Thus, dedicated two-stage AC-DC converters are essential for integrating batteries into the grid infrastructure. These converters typically employ either single-phase or three-phase inverters paired with a Dual Active Bridge (DAB) in a back-to-back configuration, ensuring high efficiency and galvanic isolation. The conventional control scheme for such converters involves two separate PI-based control loops, one for each stage. This paper proposes a novel control strategy based on a Modulated Model Predictive Control (MMPC) algorithm for a two-stage AC-DC converter. The proposed approach utilizes a single cost-function optimization routine to simultaneously control both the grid-side and the battery-side stages, simplifying the control architecture and enhancing performance. Finally, the proposed MMPC is implemented on a DSP and FPGA-based board and the experimental results show that the proposed solution achieves faster dynamics than conventional control with no overshoots or undershoots.

This paper presents a charging architecture for the Reconfigurable Cascaded Multilevel converter, which was specifically designed for electric vehicle (EV) powertrain applications. The RCMC topology is capable of executing power conversion and actively managing battery systems concurrently. The active battery management is achieved using the Reconfigurable Battery Module, which regulates the serial connection of cells via a switch pattern. In this paper, the RCMC is directly interfaced with an AC three-phase power system, facilitating the dynamic control over battery cells charging. Its inherent design allows for the implementation of various charging algorithms, customizable to specific requirements, without necessitating additional intermediary power stages. Firstly, an overview of the RCMC topology is given, and an analysis to define the optimal filter inductance is carried out. Subsequently, after the AC system characteristics are explained, two charging algorithms are presented and described: one prioritizes State of Charge (SOC) balancing among battery cells, while the other focuses on minimizing power losses. Moreover, a time estimation computation for the RCMC is carried out considering a two-level AC charging station. The result is compared with the time required for a conventional battery pack. The results show a reduction of 10 s in charging time for a mere 20% increase in SOC. Finally, the experimental setup is presented and used to validate the efficacy of the proposed algorithms.