Fei Xue

This paper demonstrates a monolithically integrated current reference and current mirror based on p-GaN gate HEMT technology, designed for high-temperature applications. The p-GaN current reference is composed of one D-mode and two E-mode devices. The generated reference current is independent of supply voltage since the proposed circuit incorporates a bias circuit capable of providing a supply-voltage-insensitive bias voltage. Moreover, under the zero-temperature coefficient (ZTC) bias voltage condition, the variation in the generated reference current at 200 °C is reduced by 15.4%, compared to a conventional p-GaN current reference with a bias voltage of 5 V. Experimental results indicate that the generated reference current slightly reduced from 2.53 to 1.70 mA over a broad temperature range of 25−200 °C. In addition, a current mirror circuit based on p-GaN HEMT technology was designed to imitate a reference current. The influence of temperature on the output current of the current mirror is mitigated, which could be realised by biasing the gate-to-source voltage at the zero-temperature coefficient voltage. This design sustains the current mirror mismatch error with small variation across a temperature range from room temperature to 200 °C. These results indicate that the GaN current reference and current mirror under zero-temperature coefficient bias voltage can ensure stable output current across different temperatures, facilitating the application of fully GaN integrated circuits in high-temperature environments.

In this work, a high-performance ultraviolet (UV) photodetector (PD) based on p-NiO/AlGaN/GaN structure is proposed for optical switching or UV detection. With its good photosensitivity, a high photocurrent of 39.75 mA/mm is achieved with UV illumination. Moreover, due to the depletion of the two-dimensional electron gas by the p-NiO layer, the dark current is reduced to 5.85 × 10^-6 mA/mm. Hence, a high photo-to-dark current ratio of 6.79 × 10⁶ is achieved in this work. Furthermore, the voltage output signal is generated through monolithically integrated two-dimensional electron gas resistors, achieving a rise time τ_rise of 0.86 ms and a fall time τ_fall of 1.33 ms, respectively. This output can be directly interfaced with logic circuits without requiring an additional conversion circuit. The monolithic integration design reduces parasitic effects, thereby enhancing operating frequency. These results indicate the suitability of the proposed GaN PDs for high-sensitivity UV detection.

This paper focuses on the modeling, analysis, and mitigation of a new false data injection attack targeting the virtual output impedance control of droop-based distributed energy resources in a microgrid with distribution lines of low X/R ratios. It is shown for the first time that the virtual output impedance-targeted false data injection attack is capable of destabilizing the autonomous microgrid, especially when the distribution lines' X/R ratio is sufficiently low. Through small-signal modeling, the low-frequency oscillations and stability margin are comprehensively characterized, providing a theoretical basis for designing resilience-enhancing virtual output impedance controls against false data injection attacks. This method can provide additional damping to microgrids with any low X/R ratio, ensuring an adequate stability margin. To simplify the small-signal model for design analysis, the work adopted the assumptions of light and high-power factor loads, as well as small voltage-angle variations across the microgrid network. A design guideline that considers the damping ratio and the diminishing-return nature of the stabilizing effect provided by the virtual inductance has been developed. The proposed modeling, attack analysis, and resilience-enhancing techniques are comprehensively demonstrated and validated on a two-DER microgrid through simulations and experiments. It is shown that the enhanced virtual output impedance control successfully improves the resilience of the microgrid against false data injection attacks, suppressing low-frequency oscillations and maintaining grid stability.

This study introduces a coordinated two-level scheduling framework for networked multi-microgrid systems, integrating day-ahead planning and real-time control under a “day-ahead guided real-time optimization” paradigm. The core innovation lies in harmonizing long-term planning with real-time adaptability to enhance both cost efficiency and system robustness. At the day-ahead level, this work proposes a hybrid algorithm that combines State-Adversarial Deep Reinforcement Learning (SA-DRL) with the Alternating Direction Method of Multipliers (ADMM). SA-DRL enables each microgrid to learn robust scheduling strategies in the presence of potential state disturbances, while ADMM ensures distributed coordination among microgrids for optimal energy exchange. At the real-time level, control policies derived from the trained SA-DRL models are refined using a dynamic penalty mechanism, aligning real-time actions with day-ahead plans to reduce forecast-induced uncertainty. Furthermore, an extended Long Short-Term Memory (xLSTM) network is employed for multi-feature rolling prediction of renewable generation and load demand, supporting both scheduling levels with more accurate temporal forecasting. By integrating prediction, learning, and optimization within a two-level architecture, the framework enhances resilience against data noise, cyber-physical threats, and temporal coupling. Experimental evaluations confirm that the proposed method reduces operating costs and improves both control stability and system robustness.

A monolithically integrated voltage reference based on p-GaN HEMT technology is demonstrated in this work. The proposed two-stage structure can improve the stability of the generated reference voltage over a wide range of the supply voltage and temperature. The static and dynamic performance was measured at various temperatures. Experimental results indicate that the output voltage is stable at 1.3 V when the supply voltage rises from 2.8 V to 40 V, with a line sensitivity of 0.035 %/V at room temperature. When the measurement temperature increases to 250 ∘C, the generated reference voltage slightly decreases to 1.25 V with a temperature coefficient of -22.1 ppm/∘C. The power supply rejection ratio of this work is competitive, as the power supply rejection ratio changes from -46.64 dB to -56.2 dB, in which the noise frequency varies from 10 Hz to 5 MHz. The voltage variation of the generated reference voltage is relatively small when the frequency exceeds 5 MHz. The results show that the proposed work is particularly suitable for all-GaN monolithic integration circuits that require thermally stable bias voltages with high immunity to the supply voltage variation.

High-performance motor drives rely on closed-loop controls that typically obtain rotor position from a physical encoder or a rotor position estimator. However, it is well established that there may be discrepancies between the measured/estimated position and the actual one. This may be due to the loosening of the encoder's mechanical fixing, initialization errors, sensorless estimation errors, etc. The rotor position error, if left uncompensated, may lead to torque fluctuation and reduced system efficiency. Different from the mainstream iterative or model-based methods introduced thus far, this paper focused on a data-driven solution that is based on the use of lightweight Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks, realized in conjunction with real-time embedded microcontrollers. Benchmarked against the familiar choice of Multilayer Perceptron (MLP), these emerging recurrent neural networks (RNNs), which have received tremendous attention in computer science subjects but much less in power electronic-based electric drives, are designed for estimating position errors with high accuracy. Upon careful consideration of the embedded data in the stationary and rotating reference frames, data down sampling, and real-time computing capability, this paper shows that these emerging RNNs are potentially more robust against measurement noises and harmonics inherently present in drive systems. They are proven to better generalize to non-training operating points or data, constituting an essential feature when dealing with closed-loop control's experimental data. The proposed lightweight LSTM- and GRU-based neural networks are extensively validated using a 2.2 kW interior permanent magnet synchronous motors (IPMSMs) through simulations and experiments for estimating the step- and ramp-type dynamic rotor position errors. The comparative evaluation against the classical iterative rotor position correction method confirms its superiority in terms of estimation speed and accuracy, suggesting a good potential of the data-driven concept in improving electric drives.