Xun Wu

DC-link capacitors in rail transit are susceptible to failure under hard working conditions, and it is important to estimate the capacitance for safe operation purpose. However, current approaches struggle to obtain accurate capacitance estimations in rail applications because of noise interference and modification restriction. Therefore, an effective method is proposed for capacitance estimation of dc-link capacitors in rail transit. The variational Bayesian adaptive Kalman filtering is designed for parameter calculation, and a novel update mechanism of prior matrix is introduced in the fixed-point iteration to improve calculation accuracy and noise immunity. The capacitance can be precisely calculated by input and capacitor voltages. The proposed method was tested on several metro vehicles, and the average error was 0.74%. Its feasibility within a certain range of signal-noise ratio, voltage deviation, and resistance drift was also proved. Comparison illustrated that the proposed method has advantages in estimation accuracy and convergence stability.

In real-world electric power systems, electrical faults often result in significant economic losses, making it critical to prevent their occurrence whenever possible. Consequently, fault samples are typically scarce and challenging to acquire. To address this issue, we propose a novel generative model that enables the controlled generation of large quantities of virtual fault samples. Unlike existing Conditional Wasserstein Generative Adversarial Networks (CWGANs), we introduce a novel conditional batch normalization module. This module jointly normalizes the input noise distribution and label embeddings, ensuring stable training and precise control over the generated outputs. Furthermore, this design mitigates inter-class mode collapse, which is a common issue in conditional generation caused by significant disparities in label embeddings. In addition, we propose a TCN-SE component equipped with SE blocks for adaptive feature recalibration, which effectively captures temporal dependencies in the fault data and performs accurate fault classification. Finally, detailed experiments are conducted to evaluate the effectiveness of the proposed model. The results demonstrate that the proposed generative module achieves an average improvement of 4.46% across six different models. Additionally, the proposed TCN-SE model achieves the highest accuracy of 96.36%. This study provides a robust framework for fault data augmentation and diagnosis in power electronic systems, contributing to the reliability and maintenance of threephase T-type inverters.

Film capacitor is widely used as dc-link capacitor in rail transit converters. Its capacitance decreases under hard working conditions and the capacitance estimation is necessary for condition monitoring. Currently, there are several challenges in rail transit applications: strong noise, low sampling frequency, and modification. In this article, a capacitance estimation method based on maximum likelihood and variable convergence factor is proposed. The input and capacitor voltages are used for modeling. The initial values of model parameters and signal noise are evaluated by recursive least square (RLS). Then, the maximum likelihood (ML) with Newton–Raphson iteration is used for parameter calculation, and a variable convergence factor is designed to solve the numerical divergence problem. The capacitance is accurately estimated and the challenges are well addressed. The proposed method was verified on the vehicles of Shenzhen Metro Group, and the error was within 2%. Performance analysis showed that the proposed method had certain immunity in the case of early signal faults and resistance offsets. The comparison results proved the advantages of the proposed method in accuracy and convergence speed.

DC-link capacitor is the important device and the vulnerable electronic part of the train converter. The capacitance decline of the DC-link capacitor is accelerated because of the influence of voltage, current, charge-discharge frequency, temperature and other factors. This will lead to the circuit performance degradation and even capacitor failures. Therefore, the capacitance estimation of DC-link capacitors is necessary and of great significance to the safe operation of the train. However, the voltage sensors used in the train converter have large noise and low sampling frequency, which makes it difficult to obtain accurate capacitance estimations with existing methods. In this paper, an effective capacitance estimation method is proposed to solve these problems. The pre-charging process is analyzed and an estimation model is built to weaken the impact of the sensor noise. Based on the model, a three-stage iteration algorithm is used for capacitance estimation, which can effectively search the optimal solution at a low sampling frequency and further reduce the noise influence. The proposed method was verified on a metro vehicle and compared with other algorithms. The experimental results show that the aforementioned problems can be well solved by the proposed method and the estimation error is within 2%.

The accurate prediction of the state of charge (SOC) of Ni-Cd batteries is critical for developing battery management systems for high-speed trains. To address the challenges of the large floating charge voltage of Ni-Cd batteries and the vulnerability of a battery’s SOC to environmental factors such as temperature, this paper proposes an adaptive adjustment mechanism-based particle swarm optimization (APSO) generalized regression neural network (GRNN) model. The proposed model introduces the concept of the particle aggregation degree to quantify the convergence of the particle swarm optimization (PSO) algorithm. Furthermore, the speed weight of the particle swarm is adaptively adjusted using a comprehensive loss function to optimize the parameters of the GRNN model. To validate the proposed method, simulation experiments are conducted under test conditions using Ni-Cd batteries, and the prediction accuracies of various algorithms are compared. The experimental results demonstrate that the APSO-GRNN model significantly reduces the model’s prediction error. In addition, under the influence of different temperatures and noises, this method demonstrates strong robustness and high practical application value by accurately predicting the SOC, even with limited data samples.

Passing the neutral section can causes transient impact and speed loss of the train. Therefore, whether the potential danger of the pantograph-catenary system can be eliminated under the condition of full load operation becomes one of the essential factors for the safe and efficient operation of electric locomotives. This paper proposes an advanced method to optimize the structure and control algorithm, according to the drawbacks of existing methods, such as miniature circuit capacity, poor output driving ability, low control accuracy, and slow system response speed. In this method, the pantograph-catenary arc is restrained by controlling the dynamic alternating transfer of instantaneous power. Besides, the voltage pre-modulation algorithm is used to realize flexible voltage switching between different adjacent power supply arms. Compared with the existing methods, this method can provide better system response output capability. And this can effectively suppress transient impact and ensure full load train operation safety. Finally, in terms of transient suppression effect and resistance to external environmental interference, a series of functional verification and horizontal comparison experiments have proved the feasibility, superiority and robustness of this method.

DC-link capacitor is one of the significant parts of traction converters. Due to the impact of electrical stress, high temperature and humidity, its capacitance degrades faster than expected. Accurate capacitance estimation is critical to the condition monitoring of DC-link capacitors, which is also the basis of online capacitor life prediction. Currently, there is a big challenge for capacitor capacitance estimation in railway applications. The noise fluctuation of the voltage sensor may be nearly equal to that of ripple voltage, leading to the considerable errors of the existing capacitor condition monitoring methods. Therefore, a capacitance estimation method based on pre-charging model and noise evaluation is proposed in this paper. The capacitance is estimated by improved recursive extended least square (RELS). The influence of the noise can be effectively reduced and both the calculation accuracy and convergence speed can be well guaranteed through an independent noise estimation model. The proposed method has good immunity in a certain range of signal-noise ratio (SNR) or signal bias. It shows good performance in experiments.

Open-circuit (OC) faults of inverter switch, clamping diode and load are easily misdiagnosed by most of the existing methods due to the extremely similar fault features. In this regard, a diagnostic method based on line voltage over-modulation and amplitude features is proposed in this paper to accurately locate coupling faults of three-level neutral-point-clamped (3L-NPC) inverter and load. Only two line voltages are used as diagnosis variables, and additional sensor is not required. The proposed diagnosis method consists of two parts: fault detection and fault classification. The fault detective algorithm only requires a small amount of computation and is suitable for uninterrupted execution. The classification algorithm starts immediately when the fault is detected, and the over-modulated voltage features are used to locate the fault. The proposed method can usually complete the diagnosis in 1/4 of fundamental period. The proposed method is also robust to different modulations, carrier frequencies and load variations. Experiments demonstrate that the proposed method can distinguish the OC faults of semiconductors (including switches and clamping diodes) and load effectively.