Zhan Li

Switching harmonics generated by power electronic converters have become a critical power quality issue in more-electric aircraft (MEA) dual-generator power systems, and accurate harmonic prediction is essential for effective harmonic cancellation. However, conventional analytical models often suffer from limited prediction accuracy, while data-driven methods are constrained by the scarcity of labeled data under certain operating conditions. This paper presents a method for harmonic prediction and cancellation in the dual-generator power systems of more-electric aircraft. To address the accumulation of switching harmonics on the DC bus, a comparative study is conducted among simplified analytical models, feedforward neural networks, and transfer learning-based models. A harmonic modeling framework and phase shift optimization strategy are established for harmonic cancellation. To overcome data limitations on the high-pressure shaft, transfer learning is employed to transfer knowledge learned from the low-pressure side to the high-pressure side. Results show that data-driven methods outperform traditional models, while transfer learning-based models further improve prediction accuracy and generalization under limited data conditions.

The triple active bridge (TAB) enables three-port bidirectional transfer with high-frequency isolation, making it highly suitable for electric aircraft demanding efficient and lightweight power systems. However, TAB control is challenged by inherent port coupling, where disturbances at one port propagate throughout the system, and meanwhile current stress and efficiency constraints further complicate the control design. Recent advances in deep reinforcement learning (DRL) enable adaptive, model-free control but face sim-to-real limitations. To address these challenges, this paper proposes a physics-informed online-trained DRL (OT-DRL) framework for TAB converters: physical constraints are embedded into the reward function and action space, while real-time hardware data is integrated into the replay buffer to bridge the sim-to-real gap. The proposed online-trained DRL approach improves learning efficiency, transient dynamics, and ensures zero voltage switching under all conditions. Experiments demonstrate 11.3% higher episode efficiency, 2.6% greater final reward, and significant enhancements in dynamic response, 15% current stress reduction, and overall efficiency across various operating scenarios¹.

The dual active bridge (DAB) converter has become a popular isolated solution to integrate energy storage systems (ESSs) and dc microgrids (MGs). However, constant power loads (CPLs) and pulsed power loads (PPLs) may reduce system damping and cause voltage oscillations in DAB converter-based ESSs (DAB-ESSs). An artificial neural network-based active disturbance rejection control (ANN-ADRC) is proposed to regulate constant output voltage quickly and accurately under different operating conditions. Firstly, the ADRC controller is designed based on the small-signal modeling of the DAB-ESSs. Feedforward compensation and uncertainty estimations of the extended state observer (ESO) help to improve the dynamic performance and to reduce the number of current sensors. Then, after satisfying the conditions of stability analysis, the parameters of the ADRC controller are selected automatically via ANN. The ANN is trained with two inputs (ADRC controller parameters) and two outputs (performance indicators of the ADRC controller). The well-trained ANN can be used as a surrogate model to obtain the optimal solution of the objective function easily and quickly. The proposed ANN-ADRC algorithm with selected parameters is implemented and validated on the hardware experimental setup. The experimental results illustrate that the proposed controller can achieve fast dynamic performance under various operating conditions.

This article proposes a current path tracking approach that uses a current path state to correlate device status and operating mode of an inverter under normal and abnormal conditions. Based on graph theory, a node-path modeling method is proposed and applied in the T-type inverter topologies. The model reflects system state transitions, given healthy or faulted transistors. Effective conduction paths and node potentials can be derived with the model to indicate circuit faults. In addition, a hierarchical diagnosis method is proposed for single-transistor open-circuit (OC) faults based on estimated output voltage vectors. The diagnosis involves two steps. First, an OC fault can be detected by following a residual between an actual voltage vector and its reference voltage vector calculated from duty ratios of intended current paths. Next, faulty transistors sharing similar features can be distinguished by adopting a current path state reconfiguration approach. Diagnosis variables can be obtained from the controller, and their errors, because of sampling, signal delay, and dead time, are taken into account to avoid misdiagnosis. Experimental results verify the robustness and effectiveness of the proposed method.

To combine the advantages of both model-driven and data-driven methods, this paper proposes a model-data-hybrid-driven (MDHD) method to diagnose open-switch faults in power converters. This idea is based on the explicit analytical model of converters and the learning capability of artificial neural network (ANN). The process of the method is divided into two parts: offline model analysis and learning, and online fault diagnosis. For both parts, model-driven and data-driven are combined. With the model information and data-based learning capability, a fast diagnosis for various operating conditions can be achieved without high computation burden, tricky threshold selection and complex rulemaking. This can greatly contribute to the practical application. The open-switch fault diagnosis in a two-level three-phase converter is studied for method validation. For this converter, an ANN is trained with two input elements, seven output elements, and two neurons in the hidden layer. Experimental results are given to demonstrate good performance.

The open-switch fault of voltage source converters (VSC) may change the switching state and distort the current, which is the basis for model-based fault diagnosis methods. Based on the analysis of the switching state, this letter presents a post-fault current model to estimate the current of each phase after one open-switch fault occurs. Combined with the current model-based method, the proposed post-fault model can be used not only to check the accuracy of the diagnosis result but also to detect multiple open-switch faults. The current, voltage and driving signals used in this model are all from the VSC control system, therefore, this model can be easily implemented without additional sensors or hardware. Finally, the effectiveness of the proposed model is verified by the experimental results.

A redundant unit has been utilized in prior work for a T-type three-level inverter to handle multiple open-circuit faults, but the output was limited to two levels. In this paper, equivalent switching state and improved virtual vector methods are proposed to complement switching states and provide better fault-tolerant performance. Under fault scenarios, equivalent replacement of unreachable switching states can be used to generate the target line-to-line voltages. If the replacement cannot be achieved by equivalent switching states, a virtual vector with redundant capability is employed to support two and three-level modulation at the output. The rearranged switching sequences can be formed by employing virtual vectors and existing vectors. In addition, by reconstructing the virtual vectors, a balancing method is adopted to eliminate neutral point voltage offset in cases of multiple faults. Experimental tests verify the effectiveness and feasibility of the proposed method.

This article considers fault detection in the case of a three-phase three-wire (3P3W) inverter, when only two current sensors are used to save cost or due to a faulty current sensor. With two current sensors, there is no current method addressing the diagnosis of both IGBT open-circuit (OC) faults and current sensor faults. In order to solve this problem, this article proposes a method which innovatively combines two kinds of diagnosis variables, line voltage deviations and phase voltage deviations. The unique faulty characteristics of diagnosis variables for each fault are extracted and utilized to distinguish the fault. Using an average model, the method only needs the signals already available in the controller. Both IGBT OC faults and current sensor faults can be detected quickly in inverter mode and rectifier mode, so that the converter can be protected in a timely way to avoid further damages. In addition, error-adaptive thresholds are adopted to make the method robust. Effects such as system unbalance are analyzed to ensure that the method is robust and feasible. Simulation and experimental results are used to verify and validate the effectiveness of the method.