Search

The distribution network with high penetration of renewable energy such as wind and photovoltaic power has higher flexibility and power supply efficiency, but it also faces more faults and uncertainties. Traditional dynamic reconfiguration under fault conditions are still limited by problems such as low load recovery rate and strong decision conservatism. To overcome these challenges, this article proposes a dynamic reconstruction strategy for distribution network under fault conditions that takes into account multivariate uncertainty. Firstly, in response to the uncertainty of distributed power generation output and load demand in the distribution network, an interval prediction method is adopted to construct a uncertainty model for source and load side. Then, the Latin hypercube sampling method is used to generate multiple operation scenarios, and computational efficiency is improved by reducing scenario samples using Cholesky sorting principle and synchronous backpropagation reduction method. Finally, a robust dynamic reconstruction model based on mixed-integer second-order cone programming (MISOCP) is constructed, and the feasibility and robustness of the proposed dynamic strategy are verified using the improved IEEE-33 node system. Through analysis, the proposed method effectively addresses the risk factors in the operation, thus improving the safety and reliability of the distribution network.

The use of computer vision to estimate the Body Condition Score for cow has demonstrated to be feasible. However, most research has been limited to fixed camera positions, which restricts the technique’s usefulness. This research acquired cow data at various distances and angles to investigate the impact of distance and different depth images restoration method on scoring accuracy. A U-Net neural network-based model was employed for background segmentation. The model proposed in this study performs best among SegNet, UNet, UNet++ and DeeplabV3, with significant performance improvement and superior overall segmentation accuracy, and has stronger foreground object identification capability. Additionally, we utilize the concept of human feature point localization to pinpoint the positions of cow feature points. The results show that compared to Hourglass, CPN, and Hrnet, the model in this study has significant advantages in three core indicators: accuracy, recall, and mAP. Moreover, we presented an unsupervised depth image reconstruction model based on the Denoising Diffusion Probabilistic Model and Unet++ Mode, facilitating the measurement and scoring of cow characteristics. Finally, the measurement results of cow body size using the Lerp model, autoregressive model, GAN model, and the depth image completion model proposed in this study were compared. The method proposed in this study was found to be effective and feasible for meeting actual production requirements at a camera distance of 1–2 m from the cow, achieving high accuracy with a coefficient of determination above 0.9. Additionally, the three models used were effective in handling small-scale depth deficits, with a high degree of agreement between machine and manual measurements. The DDPM-based depth image reconstruction model was found to produce the most accurate results when the camera distance from the cow was between 1-3m. Therefore, this research makes it possible to make flexible measurements in the near range using existing methods. Accurate measurements at longer ranges are influenced by depth image quality and feature point localization accuracy, which require more advanced techniques for further investigation.

As a self-supervised learning method, the graph contrastive learning achieve admirable performance in graph pre-training tasks, and can be fine-tuned for multiple downstream tasks such as protein structure prediction, social recommendation, etc. One prerequisite for graph contrastive learning is the support of huge graphs in the training procedure. However, the graph data nowadays are distributed in various devices and hold by different owners, like those smart devices in Internet of Things. Considering the non-negligible consumptions on computing, storage, communication, data privacy and other issues, these devices often prefer to keep data locally, which significantly reduces the graph contrastive learning performance. In this paper, we propose a novel federal graph contrastive learning framework. First, it is able to update node embeddings during training by means of a federation method, allowing the local GCL to acquire anchors with richer information. Second, we design a Self-adaptive Cluster-based server strategy to select the optimal embedding update scheme, which maximizes the richness of the embedding information while avoiding the interference of noise. Generally, our method can build anchors with richer information through a federated learning approach, thus alleviating the performance degradation of graph contrastive learning due to distributed storage. Extensive analysis and experimental results demonstrate the superiority of our framework.

Nonlinear control of light-matter interaction plays an important role in various optical phenomena, including spin-orbit interaction of light. In this paper, we demonstrate that the spin-orbit interaction of graphene-wrapped nanoparticles can be tuned by the input intensity of light. Typical optical bistable responses are observed in the scattering efficiency and the local electric fields. In the optical bistable regime, the strength of spin-orbit interaction in the near-field is significantly tuned in the switching-up and switching-down branches. The tunability of spin-orbit interaction enables promising nonlinear control schemes in nanophotonic devices.

To comprehensively evaluate the camouflage effect of a camouflaged vehicle in a complex background based on the detection and identification processes of the time-limited search model, the visual saliency degree and feature similarity degree were used to represent the camouflage effect, and a camouflage effect evaluation method was proposed. For the visual saliency degree algorithm, in which the concepts of local saliency degree and target saliency degree were introduced, the visual saliency features of the camouflaged vehicle and background image were manually extracted, multiscale saliency matrices were constructed, and the comprehensive saliency function was used to calculate the visual saliency degree. Based on the bag of visual words model, the feature similarity degree algorithm extracted the local features from public images of the camouflaged vehicles, clustered these features to form a visual word bag, and calculated the feature similarity degree by comparing the local features of the camouflaged vehicle with the visual word bag. Through experimental verification and analysis of the results, the method considering complex landform background characteristics, battlefield target reconnaissance processes, and different camouflage statuses of the vehicle could objectively obtain the camouflage effect value of vehicles, quantitatively determine the advantageous application environment of different camouflaged states, and provide data support for camouflaged state design and concealed area selection.

To improve the efficiency for managing power systems with constant time delay, an improved frequency-weighted model reduction is proposed. First, the Padé approximant allows to represent the time delay in the state-space form and build a linear model of the closed-loop power system containing a wide-area damping controller. Then, a weight function is introduced to solve the equilibrium transformation matrix and balance the original system. Finally, the residual order reduction is applied to the model of the power system considering time delay. The proposed method is evaluated in the two-area four-machine test system and New England system, and the effectiveness of the proposed reduction method is verified.

Vehicle tracking based on surveillance videos is of great significance in the highway traffic monitoring field. In real-world vehicle-tracking applications, partial occlusion and objects with similarly appearing distractors pose significant challenges. For addressing the above issues, we propose a robust multivehicle tracking with Wasserstein association metric (MTWAM) method. In MTWAM, we analyze the advantage of the 1-Wasserstein distance (WD-1) on partial occlusion and employ the WD-1 as the similarity criterion to measure the similarity between tracklets and detections. Moreover, for distinguishing different objects with a similar appearance, we improve the feature presentation of vehicles by developing target-specific feature sparse coding (TSSC). To demonstrate the validity of this method, we present a quantitative evaluation of both the UA-DETRAC dataset and our vehicle highway surveillance videos dataset (VecHSV). In both cases, our method achieves state-of-the-art performances.

As the deployment of wide-area measurement systems (WAMS) expands, data-driven methods are playing an increasingly important role in transient stability assessment (TSA). However, if the measured data is disturbed by noise or the topological structure of the power system changes, the performance of the model based on data-mining would decline so that it could not meet the needs of real-world scenarios. In this paper, we develop a TSA methodology based on extreme gradient boosting (XGBoost) and factorization machine (FM). Utilizing XGBoost to complete automatically the feature construction and transform the power flow into a sparse matrix. The influence of noise can be reduce due to the sparsity of the new feature set. On this basis, FM algorithm, which has advantage in processing the large sparse matrix, is introduced into the model to complete fault classification. Furthermore, the feature crossing function of FM further mines the interactive information of the spatio-temporal features. For changed topology scenarios, we propose a extended-training set scheme. Adding some pivotal data of topology changes to the training set improves the robustness of the model. Compared with existing studies, the proposed assessment model based on XGBoost-FM not only has the better generalization performance in the case of noise interference or changed topology, but also has the least time consumption and complexity.

With the development of renewable electric power generation and ultra-high voltage direct current (UHVDC) transmission, the transient overvoltage in the sending end electric power system with UHVDC occurs, but it is lack of quantitative transient voltage assessment index (TVAI) to assess the overvoltage. To deal with the above problems, this paper proposes a new quantitative transient voltage assessment index considering overvoltage and voltage dip in the sending end electric power system with UHVDC, and applies it to analyze different influencing factors of overvoltage and determine the dynamic reactive power compensation capacity. In detail, first, according to the principle of the transient voltage assessment index considering voltage dip and the overvoltage assessment index for the distribution network, a new transient voltage assessment index considering transient overvoltage in the sending end electric power system with UHVDC is proposed with the approximated step function. Then, based on the requirements of the system with UHVDC, the parameters of the proposed transient voltage assessment index are determined. Furthermore, with the absolute and relative sensitivity, the impact of the parameters of voltage dependent current order limiter and current control amplifier on the transient overvoltage is analyzed and compared quantitatively. Finally, the impact of reactive power compensation equipment, such as synchronous condenser, static var compensator, and static var generator, on the transient overvoltage of the system with UHVDC is compared quantitatively, and the reactive compensation capacity needed to maintain transient voltage security and stability is estimated. The two applications show the effectiveness and advantages of the proposed index in the sending end electric power system with UHVDC.

The method based on the two-stream networks has achieved great success in video action recognition. However, most existing methods employ the same structure for both spatial and temporal networks, leading to unsatisfied performance. In this paper, we propose a spatiotemporal heterogeneous two-stream network, which employs two different network structures for spatial and temporal information, respectively. Specifically, the Residual network (ResNet) and BN-Inception are utilized as the base networks to present the spatiotemporal characteristics of different human actions. In addition, a segmental architecture is employed to model long-range temporal structure over video sequences to better distinguish the similar actions owning sub-action sharing phenomenon. Moreover, combined with the strategy of data augment, a modified cross-modal pre-training strategy is proposed and applied to the spatiotemporal heterogeneous network to improve the final performance of human actions recognition. The experiments on UCF101 and HMDB51 datasets demonstrate the proposed spatiotemporal heterogeneous two-stream network outperforms the spatiotemporal isomorphic networks and other related methods.

When a large amount of wind power is connected to the power grid, certain control methods need to be taken for wind turbines (WTs) so that WTs can respond to system frequency changes and maintain system frequency stability. Based on the frequency control method of a single doubly fed induction generator, a deloading power coordinated distribution method for frequency regulation by wind farms considering wind speed differences is proposed to better utilize the wind farm frequency regulation ability. A reasonable distribution of the deloading power of a wind farm for participation in grid frequency regulation to better utilize the wind farm frequency regulation ability. A quantitative calculation formula for the wind speed, total deloading power and single WT deloading power is established by using a variable parameter deloading power coordinated distribution coefficient based on the wind speed. The deloading power is distributed according to the wind speed of the WT units to adapt to the wind speed distribution differences between WTs in cluster wind farms and fully utilize the wind farm frequency regulation ability. A typical doubly fed induction generator (DFIG) wind farm is used as a simulation case. The experimental results show that the control strategy can distribute the output of each WT, improve the frequency response of the wind farm and enhance the frequency stability of the system.

12-pulse rectifier is extensively used in high power rectification, and the delta-connected autotransformer and wye-connected autotransformer are its two most popular phase-shift transformers. This paper compares the 12-pulse rectifiers using the two transformers via calculating the input line current, load voltage, kVA ratings of the two autotransformer, kVA ratings of the auxiliary magnetic devices. From the viewpoint of power quality of AC mains and DC side, the two 12-pulse rectifiers are the same. The kVA rating of the IPR in the two 12-pulse rectifiers are equal, and the kVA rating of the ZSBT in the two 12-pulse rectifier are also equal to each other, under the same load power. However, the kVA of the delta-connected autotransformer is less than that of the wye-connected autotransformer under the same load power. Some experimental results are shown to validate the correctness of the theoretical analysis.

We demonstrate two kinds of stable ultrafast fiber lasers mode-locked by topological insulator (TI) Bi2Se3 nanosheets interacting with photonic crystal fiber (PCF). The saturable absorber (SA) produced by drying TI solution filled in 5.5 cm PCF provides a modulation depth of 5%, and its damage threshold is highly enhanced due to the large evanescent field interaction length. When the SA is incorporated in an erbium-doped fiber ring cavity with anomalous dispersion, a conventional soliton with 908-fs duration is obtained, and no wave-breaking and SA damage is ever shown, even for pump power at 900 mW. Through tuning polarization controller, as high as tenth harmonic mode-locking, namely, 202.7 MHz, is obtained, and it operates stably for 8 hours under lab conditions. Dissipative soliton with a duration of 7.564 ps is generated when the dispersion of cavity changes from anomalous to normal, which can be further compressed to 245 fs by additional 15-m single mode fiber, and a maximum average output of 75 mW is obtained. This TI nanosheets-PCF structure possesses advantages of stable saturable absorption performance, high damage threshold, and freedom from special package, which make it an effective choice for a passively mode-locked laser.

A magnetic sensor utilizing long-period fiber grating (LPFG) written by high frequency CO2 laser pulses in a D-shaped fiber and the magneto-optical effect of magnetic fluid is proposed. Because of the high evanescent field of the grating structure in D-shaped fiber, by immersing the LPFG with a period of 685 μm into a water-based magnetic fluid within a capillary tube, the redshift as high as 33.5 nm of the resonant wavelength of the grating is observed when the external magnetic-field intensity is 189.7 mT, resulting in a sensitivity of 176.4 pm/mT. The hysteresis effect is observed and explained by the characters of magnetic fluid. Such kind of sensors would find potential applications in magnetic sensing fields.