Pinjia Zhang

The conventional deadbeat predictive current control (DPCC) for current-source inverter-fed permanent magnet synchronous motor (CSI-PMSM) deteriorates control performance when subjected to disturbances such as model parameter mismatch. Therefore, the model-aided extended state observer (MAESO) CSI-DPCC is proposed. First, a high precision discrete time CSI-PMSM prediction model is established, integrated with an ESO to combine model predictive capabilities with lumped disturbance estimation, thereby enhancing state variable prediction accuracy. The MAESO effectively enhances parameter robustness. Meanwhile, predicting known states through the prior model reduces the observer computational burden and increases the system control bandwidth. To eliminate the tracking error caused by disturbances, the system disturbance estimated by MAESO is fed back to the deadbeat model controller, which can estimate and compensate for internal and external disturbances in real time without requiring an additional integrator to eliminate steady-state errors. In addition, the parameter design method of MAESO-CSI-DPCC is proposed based on the Jury stability criterion for discrete systems. Finally, comparative simulations and experiments between the CSI-DPCC and the proposed MAESO-CSI-DPCC verify the effectiveness and superiority of the proposed strategy.

The MMC-HVDC system is widely used for offshore wind-power grid connection, and the reliability of such system has increasingly become a key focus in industry. Existing studies primarily concentrate on mitigating fault impacts and implementing rapid protection measures, while less attention has been given to the cable insulation, which is the primary cause of the majority of faults. To solve this challenge, an online cable monitoring scheme for MMC-HVDC systems based on undamped resonance characteristics is proposed in this article, effectively monitoring the ongoing aging process and offering valuable reference data for maintenance scheduling. The undamped resonance of MMC-HVDC system is theoretically modeled and analyzed in detail. Then, the relationship between cable insulation aging and resonance characteristics is thoroughly modeled. A novel monitoring solution using the high-precision frequency estimation algorithm is further proposed, which can assess the insulation condition of localized aging segment. Finally, extensive simulations and experiments are carried out to prove the effectiveness and robustness of the proposed scheme. Compared with existing approaches, the proposed method requires no additional equipment and is fully non-intrusive, making it highly applicable for practical implementation.

Wind power, as an important form of clean energy, relies on the safe and stable operation of energy conversion systems (ECSs). However, ECS anomalies often lead to unexpected shutdowns, posing significant risks to turbine reliability. Existing ECS anomaly detection methods generally face two major limitations: (1) a single-level detection structure that hinders hierarchical monitoring, and (2) substantial steady-state redundancy in dynamic modeling, which reduces the model’s sensitivity to dynamic anomalies. To address these issues, this study proposes a multilevel dynamic anomaly detection method for ECSs based on hierarchical fast-feature latent variables. First, a correlation-weighted block-structured slow-feature analysis is employed to enhance the dynamic representation of raw supervisory control and data acquisition data while achieving slow–fast-feature separation. Subsequently, a latent variable model is used to construct an independent dynamic space, eliminating steady-state interference introduced by slow features and extracting fast-feature latent variables (FFLVs). Finally, FFLVs from different ECS components are integrated and combined with global and local FFLV spaces, as well as the residual space, forming a multilevel dynamic anomaly detection and root-cause analysis framework spanning system, device, and feature-levels. Experimental results based on real wind farm ECS fault scenarios verify the effectiveness and advanced performance of the proposed method.

Position sensorless synchronous reluctance motor (SynRM) drives offer a reliable and cost-effective solution for electrified transportation systems. However, their position estimation accuracy degrades under uncertainty noise interference and stator resistance mismatch. In this paper, the SynRM position estimation method based on adaptive extended Kalman filtering-enhanced hybrid flux observer (AEKF-HFO) is proposed, which combines the wide speed range of HFO with the strong anti-interference ability of Kalman filtering. A flux prediction equation combining the current and voltage models is developed for designing an EKF-based flux observer, in which an online stator resistance adaptation scheme is integrated to enhance estimation consistency under the parameter variations. To alleviate the problem of position estimation accuracy decline caused by stator resistance estimation error at low speed, a structured Sage-Husa adaptation scheme is developed. The proposed method differentiates flux-and resistance-related components in the system noise covariance matrix and propagates the uncertainty of stator resistance into the flux subspace through sensitivity analysis. In addition, the stability of the proposed method is demonstrated in details. The experimental results show that the proposed method has superior position estimation accuracy and robustness under challenging conditions including noise interference, parameter mismatches, and inverter nonlinearities.

The high electric stress experienced by insulation systems today continues to pose significant challenges to the reliability of inverter-fed electrical machines (EMs) used in transportation applications. Electrical aging caused by partial discharge is one of the primary factors threatening the reliability of EMs. Moreover, compared to EMs meant for industrial applications, the insulation systems of EMs designed for more-electric aircraft may also be affected by pressure conditions. For instance, EMs used in primary flight control surface actuators may operate during the cruise phase at altitudes of approximately 12 km and pressures as low as 200 mbar. Several studies have highlighted that low-pressure environments can accelerate the electrical aging process. However, a comprehensive lifetime model accounting for both pressure and electrical stress remains unavailable. Such a model would enable machine designers to assess the suitability of insulation systems for various applications across different operating altitudes (i.e., pressure levels). In this paper, accelerated insulation aging tests are conducted under twelve distinct combinations of pressure and voltage conditions, and a multi-stress (i.e., electrical and pressure stress) lifetime model applicable in the range 200÷1000 mbar is developed, accounting for the impact of pressure on inverse power law’s parameters.

Conventional vibration-based diagnosis methods for slight gear faults struggle with weak fault vibration characteristics and low signal-to-noise ratios. The nonlinear characteristics of gears are highly sensitive to faults. In order to accurately and timely diagnose slight gear fault, this paper proposes a novel online diagnosis method for slight gear faults by identifying nonlinear characteristics through motor multi-frequency torque injection. The proposed approach actively injects optimized multi-frequency torque signals via the motor, enabling real-time identification of second and third order nonlinear kernels from torsional responses. Experimental results demonstrate that single-tooth faults significantly increase the second-order nonlinear kernel, while uniform wear faults amplify the third-order nonlinear kernel, enabling clear fault classification. The method offers high sensitivity to slight faults and robustness under varying operating conditions, outperforming conventional vibration spectrum analysis.

Due to the increasing DC bus voltage, electrical machines (EMs) for transportation applications are exposed to a great partial discharge (PD) risk nowadays. Thus, PD risk assessment is essential for fulfilling the reliability requirement. PD inception voltage (PDIV) measurement under sinusoidal voltage is sometimes preferred because it demands an easy experiment setup, both in terms of the voltage source and PD detection system. However, the obtained results may lead to an underestimation of PD risks since, in actual applications, EMs are fed through inverters and pulsewidth modulation (PWM) voltage waveforms are applied. In this article, the difference between PDIV results obtained under sinusoidal and pulse voltage signals is quantified, and different environmental conditions are investigated, considering the potential transportation electrification applications. Finally, based on the experimental results, a stress voltage waveform enhancement factor (EF) is proposed and precisely determined. The study reveals that this factor is not constant under varying environmental conditions, with a maximum value of 1.18 observed at $220~^{\circ }$ C and 200 mbar. Furthermore, the influence of the statistical processing method on the derived factors is analyzed, showing a maximum variation of approximately 10% when different Weibull percentiles are employed.