Huimin Huang

Fast and accurate diagnosis is essential to fault ride-through and reliability enhancement of power conversion systems. For three-level neutral-point-clamped (3L-NPC) power converters, most diagnosis methods fail in distinguishing faulty insulated gate bipolar transistors (IGBTs) and diodes. Here, we propose a unified fast open-circuit fault diagnosis technique for all power devices in a 3L-NPC converter. The proposal is realized simply in two steps via the classical “quasi-power” (QP) concept of Tellegen’s theorem. First, the faulty outer IGBTs and freewheeling diodes will be allocated with one sampling period solely via comparing the QP values. Afterward, a voltage signal will be injected via manipulating the IGBT gate signals in a direct predictive control frame, and the correspondent QP value will be recalculated to distinguish and allocate the faults between an inner IGBT and a clamping diode, with one additional sampling period. Since the QP depends solely on the topological and switching state of power converters, the proposal hence immunes to both parameter and operation condition variations. Finally, both signal hardware-in-the-loop and power circuit experimental data validate the effectiveness of the proposed method at different operation scenarios.

As wind farms connect with power grids via long-distance transmission lines and multiple transformers, the resulting networks exhibit inherently weak and fluctuating grid strength—subject to dynamic variations from external disturbances like generation tripping; thus, ensuring stable operation presents a critical challenge. To address this issue, grid-following/grid-forming mode switching has emerged as a strategic approach, where wind turbine generators strategically switch between grid-following control in strong grids and grid-forming control in weak grids. This paper proposes an impedance-based switching criteria design methodology, ensuring system stability under dynamic grid strength variations. Leveraging sequence impedance analysis in frequency domain, we establish a stability boundary for mode transitions. Simulation results demonstrate that the proposed criteria maintain stable operation, enhancing robustness against grid strength fluctuations.

The parallel operation of converters is increasingly being adopted to meet higher power demands and improve reliability. However, open-circuit faults in semiconductor switches must be located quickly to prevent unnecessary derating and costly offshore interventions. In parallel topologies, zero-sequence circulating current (ZSCC) distorts phase current residuals at fault inception, making conventional residual-based fault diagnosis methods unreliable. This paper proposes a fast fault diagnosis method that integrates phase current residual analysis with real-time ZSCC polarity. The method monitors which residuals exceed a fixed threshold and compares the polarity of the dominant residual with that of the ZSCC. In this way, it can distinguish and locate single-switch, same-converter dual-switch, and cross-converter dual-switch faults. A three-stage timing window mechanism captures residual and ZSCC signatures within only a few sampling periods, enabling precise fault location without additional sensors. Experimental and simulation results verify the accuracy and robustness of the proposed method, demonstrating its potential for practical deployment in offshore wind turbines.

With the proliferation of wind power generation, the receiving end grids exhibit unprecedented dynamic characteristics, imposing critical stability challenges on grid-connected wind turbine’s converter. To address this, wind turbine converter control strategies have evolved beyond traditional grid-following (GFL) methods to include grid-forming (GFM), mode-switching, and hybrid GFL-GFM controls. This paper establishes a small-signal model for hybrid GFL-GFM-controlled wind turbines to analyze stability at varying grid strengths, guiding the selection of coefficients in hybrid mode. Simulation tests validate the theoretical framework.

Power rating of offshore wind turbines is becoming increasingly larger and their fault-ride-through capabilities are essential to enhance system reliability. Parallel power converters are commonly used to fulfill the required large power capacity. However, classical fault detection methods for the unparalleled topology usually cannot be applied herein due to the presence of zero-sequence circulating current (ZSCC) in the underlying parallel power converters, which can obscure fault characteristics. In this work, we first reveal that ZSCC induces fluctuations in phase current residuals when an open-circuit fault occurs, rendering the traditional residual threshold-based fault diagnosis ineffective. Then, we derive that the ratio of fault-to-healthy phase current residuals remains approximately at a fixed proportion despite ZSCC disturbances. Using this feature, we propose a fault diagnosis method for parallel power converters in wind turbine systems. The method achieves dynamic residual ratio extraction through sliding-window root-mean-square normalization and convolutional signature enhancement, resulting in self-adaptive threshold calibration and a very short detection time of several sampling cycles. Experimental and simulation data verify the effectiveness of the proposed diagnosis method at various operation scenarios.