To address the DC-side inter-phase voltage imbalance in star-configured cascaded H-bridge static var generator (SVG) under grid unbalanced faults in new power systems, a ripple suppression method combining a hybrid-frequency topology with triple frequency zero-sequence voltage injection is proposed. First, the working principle and power transfer mechanism of the hybrid frequency topology are analyzed, and its overall closed-loop control strategy is modeled and designed. Second, by establishing a ripple mathematical model, a key characteristic is identified: the amplitude of the mid-frequency ripple component is significantly lower than that of the fundamental component. Based on this, the optimal zero-sequence voltage phase angle is derived, and the ripple suppression effect under different zero-sequence voltage amplitudes is quantitatively analyzed. Finally, simulations under two different triple frequency zero-sequence voltage amplitudes are conducted. The results show that the ratio of peak-to-peak ripple before and after zero sequence voltage injection is approximately 1. 1 and 1. 21 for the two cases, respectively. An experimental platform validates the system performance under unbalanced conditions, demonstrating effective DC-side voltage balancing. All results are basically consistent with theoretical derivations, verifying the effectiveness of the proposed ripple suppression method.
To address the challenge of incomplete boundary features and undetermined center coordinates caused by missing object boundaries in image recognition, this paper proposes a boundary feature topological detection method based on an enhanced Hough algorithm. Firstly, a Canny edge detection algorithm integrated with morphological processing is applied to preprocessed image data to extract edge information, ensuring edge continuity and accuracy. Secondly, the improved Hough algorithm is employed to perform boundary topology and localization on images with incomplete boundary information, effectively inferring the locations and extents of missing boundaries. Experimental results demonstrate that the proposed method achieves a mean localization error of 0. 82 mm, with 97. 92% of errors below 2 mm. The approach significantly enhances the precision and effectiveness of detecting target boundary features, accurately reconstructs missing boundaries, and achieves robust localization of targets, even under conditions of partial occlusion or boundary discontinuities.
Deep learning has shown strong capabilities in modeling nonlinear relationships and complex patterns in temperature prediction. However, it faces limitations in processing multidimensional meteorological data and capturing long-term dependencies. To tackle this issue, this study proposes a WTConv-Informer based temperature prediction method enhanced with frequency-aware channel attention. By integrating wavelet convolution with the Informer model and incorporating a frequency-enhanced channel attention mechanism, the method enables deep extraction of global features and precise identification of dominant low-frequency trends and fine grained high-frequency variations in time series. Experiments conducted on the Max Planck Institute for Biogeochemistry weather station dataset show that the proposed model outperforms traditional approaches in prediction accuracy, achieving a mean absolute error (MAE) of 0. 206 4 and a mean squared error (MSE) of 0. 085 4. The findings indicate that the proposed model delivers high accuracy and robust generalization in temperature prediction and demonstrates notable advantages in handling multivariable meteorological data.
To explore the influence of molecular chain dynamics on high-temperature breakdown performance, nano-SiO 2 particles are added to epoxy composites to modulate chain dynamics. The chain characteristics are analyzed by broadband dielectric spectroscopy, and the breakdown strength is measured. The relationship between molecular chain dynamics and high-temperature breakdown is discussed. The results show that with 1% nano-SiO 2 , stable chemical bonds are generated in the interfacial region between nanoparticles and the matrix, suppressing β-chain relaxation, reducing free volume and bulk conductance, and improving breakdown voltage by 12. 68% . At high temperature, the crosslinked structure reduces the thermal expansion coefficient, further decreasing free volume compared to pristine epoxy, and increasing breakdown voltage by 16. 40% . The results lay theoretical foundation to improve breakdown voltage of polymeric materials at different temperatures by suppressing molecular chain dynamics.
Retinal fundus vessel segmentation technology is of great significance for the diagnosis and prevention of diabetes and retinal diseases. However, pixel-level annotation of blood vessels is very time-consuming and costly. In the case of limited labeled data, blood vessel segmentation remains a challenging task. To address the above problems, this paper proposes a Dual Uncertainty guided Dynamic-competitive Collaborative Mean Teacher framework ( DU-DCMT) for semi-supervised vessel segmentation. The framework utilizes a Dual Uncertainty Region-processing Module (DURM) to expand data distribution, enhance the confidence of pseudo-labels, and guide network attention to correct potential error-prone regions. Additionally, a dynamic competitive teacher model is designed to select high-quality pseudo-label generators, jointly train networks, and preserve network diversity for better segmentation performance. Experimental results show that when the labeling ratio is 0. 2, the sensitivity and accuracy of this method on the DRIVE dataset are 0. 817 3 and 0. 967 3 respectively, and on the STARE dataset are 0. 768 2 and 0. 978 3 respectively. Therefore, DU-DCMT effectively alleviates the problem of data label scarcity, and its segmentation performance outperforms current state-of-the-art methods.
