Search

For physically understanding the dynamic response of the photovoltaic (PV) integrated power system for electromechanical oscillations damping, firstly, this paper develops the linear mathematical model of a single machine infinite bus system integrated by the PV grid-tied inverter under the electromechanical time scale. Then, based on the electric torque analysis method, and the functional route of the PV grid-tied inverter variable active power control (P-control) and reactive power control (Q-control) with the power angle or grid frequency feedback as the input control signal, the key factors and influence laws of the integrated system inertia, damping level and synchronization capability are investigated. The developed model shows that enlarging the proportional-integral-derivative (PID) controller parameters of P and/or Q-control loop can effectively change electromagnetic power distribution to suppress the imbalance power, and equivalently improve the system damping level, synchronization capability and inertial effect, respectively, contributing to actively adjust the grid frequency oscillation amplitude, period and recovery speed to maintain the system control stability according to the grid code demand. Besides, under the same inverter capacity constraint, P-control mode is proved more advantageous than Q-control mode to damp electromechanical oscillations. Finally, the correctness of the physical mechanism analysis is verified by MATLAB simulations and RT-LAB experiments.

The heterogeneity of cancer reflects the complexity of genetic mutations. Dissecting the heterogeneity plays an important role in the field of biomarker discovery, targeted therapy and drug designing. As it is time-consuming to identify new biomarkers in biological experiments, various machine learning methods have been developed. However, the current methods are limited because they ignore that patients may correspond to different disease-causing genotypes. In this article, a genotype-based ensemble classifier system (GECS) is proposed which aims to explain pathologies of NSCLC from the view of genotypes and identify the genetic subtypes of tumors. The core strategy of GECS is to construct multiple independent classifiers following the principle that one classifier is constructed based on one genotype of NSCLC. The analysis of synthetic data and three microarray datasets indicated that the proposed method outperforms existing approaches in the identification of genetic subtypes of tumors. The GECS method provides a useful tool for molecular pathology researches for dissecting the heterogeneity of cancer.

Existing video domain adaptation methods focus on addressing closed set problems. However, it is nearly impossible to guarantee different domains share exactly the same set of categories in realistic scenarios. Hence, open set video domain adaptation (OSVDA) problem, which involves unknown categories, has achieved increasingly close attention. In this paper, we propose a seminal framework, which involves spatial and temporal information to address OSVDA problem. Besides, we design a novel discrimination module, i.e., Dual Metric Discriminator (DMD), to separate known and unknown categories based on implicit and explicit similarity metrics. We conduct comprehensive experiments on several benchmarks and achieve state-of-the-art performance with 40.4%, 33.7%, and 79.2% accuracy on UCF to HMDB, HMDB to UCF, and Kinetics to UCF scenarios respectively.

Compared with image-based UDA, video-based UDA is comprehensive to bridge the domain shift on both spatial representation and temporal dynamics. Most previous works focus on short-term modeling and alignment with frame-level or clip-level features, which is not discriminative sufficiently for video-based UDA tasks. To address these problems, in this paper we propose to establish the cross-modal domain alignment via self-supervised contrastive framework, i.e., spatio-temporal contrastive domain adaptation (STCDA), to learn the joint clip-level and video-level representation alignment. Since the effective representation is modeled from unlabeled data by self-supervised learning (SSL), spatio-temporal contrastive learning (STCL) is proposed to explore the useful long-term feature representation for classification, using self-supervision setting trained from the contrastive clip/video pairs with positive or negative properties. Besides, we involve a novel domain metric scheme, i.e., video-based contrastive alignment (VCA), to optimize the category-aware video-level alignment and generalization between source and target. The proposed STCDA achieves stat-of-the-art results on several UDA benchmarks for action recognition.

Modern electrical power systems have an increasing penetration of distributed generation (DG). This increase yields the requirement for integration of renewable energy and distributed energy generation into the educational curriculum of power engineering. However, the integration of associated knowledge in the curriculum framework is challenging because power system industry is experiencing a fast and continuing transition and associated teaching and learning materials are becoming quickly outdated in higher education institutions. This paper demonstrates the use of the backward design model in the power engineering discipline, where a Year 3 course is redesigned. The course is entirely implemented in an online mode, and the feedback and perception from the students are presented and analyzed. The results show the necessity of the curriculum framework development for power engineering education in a qualitative manner and provide guidance for course instructors to integrate distributed energy generation in undergraduate coursework programs.

The deployment of distributed generation contributes to the facilitation of the microgrid concept and grid modernization on a global scale. An increasing number of workforces are required for further development of microgrid systems. However, the mismatch between the University education and industry expectation places a negative impact on workforce development for future energy infrastructure. This paper introduces the implementation of a microgrid virtual laboratory for a design course. The key elements of this developed course are described in detail and the reflections from the students are investigated. This contribution also provides a statistical analysis of the students’ knowledge gap in the field of microgrids. The curricula design outlined in this study can be proposed in future power engineering education to offer competent students for the microgrid industry.

In this paper, a series-parallel hybrid dynamic voltage restorer (DVR) is taken as the research object, and the compensation strategy and control method of the DVR are proposed. The mathematical model of the current loop and voltage loop in the control strategy is simplified. Secondly, a seriesparallel hybrid topology dynamic voltage restorer is studied. This type can simplify the structure of the DVR system and improve the stability of the DVR system. Finally, the working conditions of instantaneous voltage drop and swell at different moments were simulated in MATLAB/SIMULINK. The simulation results proved that the detection method and control strategy used in the DVR in this paper have good fastness and accuracy.

By modulating the active power output of the grid-tied inverter, the photovoltaic (PV) generation system can effectively damp the electromechanical oscillation of the power grid. To analyze the influence of different PD controller gains on its output power for the typical droop and inertia control strategies, this paper takes the conventional application scenarios of PV damping electromechanical oscillation as the research example. Studies show that when the PD controller gain is amplified to the limit condition, the PV generation system will operate in the rapid power compensation (RPC) mode: based on the speed deviation polarity, the active power modulation limit (Pmppt or 0) will be used as the basis for alternate output to damp the system imbalanced power. The comparison of speed indicators shows that the RPC mode is significantly stronger than the conventional droop and inertia control for grid electromechanical oscillation damping. In view of this, this paper proposes a novel grid electromechanical oscillation damping strategy based on RPC mode, and sets the switching thresholds and corresponding active current commands for 4 operating conditions to quickly and effectively suppress the generator rotor speed oscillation phenomenon. The simulation results show that the proposed control strategy in this paper does not require a complicated control parameter tuning process, and can significantly suppress the system electromechanical oscillation, which is beneficial to ensure the power system stable operation.

The proliferation of mobile computing has prompted Pedestrian Dead Reckoning (PDR) to be one of the most attractive and promising indoor localization techniques for ubiquitous applications. The existing PDR approaches either suffer position drifts caused by accumulative errors or are sensitive to various users. This paper presents ARPDR, an accurate and robust PDR approach to improve the accuracy and robustness of indoor localization methods. Particularly, we propose a novel step counting algorithm based on motion models by deeply exploiting inertial sensor data. We then combine step counting with adaptive thresholding to personalize the PDR system for different users. Furthermore, we propose a novel stride-heading model with a deep neural network to predict stride lengths and walking orientations, thus the displacement errors are significantly reduced. Extensive experiments on public datasets demonstrate that ARPDR outperforms the state-of-the-art PDR methods.

The precise minute time scale forecasting of an individual Photovoltaic power station output relies on accurate sky image prediction. To avoid the two deficiencies of traditional digital image processing technology (DIPT) in predicting sky images: relatively limited input spatiotemporal information and linear extrapolation of images, convolutional auto-encoder (CAE) based sky image prediction models are proposed according to the spatiotemporal feature extraction ability of 2D and 3D convolutional layers. To verify the effectiveness of the proposed models, two typical DIPT methods, including particle image velocimetry (PIV) and Fourier phase correlation theory (FPCT) are introduced to build the benchmark models. The results show that the proposed models outperform the benchmark models under different scenarios.

In this paper, we design a tracking model consisting of response generation and bounding box regression, where the first component produces a heat map to indicate the presence of the object at different positions and the second part regresses the relative bounding box shifts to anchors mounted on sliding-window locations. Thanks to the resizable convolutional filters used in both components to adapt to the shape changes of objects, our tracking model does not need to enumerate different sized anchors, thus saving model parameters. To effectively adapt the model to appearance variations, we propose to offline train a recurrent neural optimizer to update tracking model in a meta-learning setting, which can converge the model in a few gradient steps. This improves the convergence speed of updating the tracking model while achieving better performance. We extensively evaluate our trackers, ROAM and ROAM++, on the OTB, VOT, LaSOT, GOT-10K and TrackingNet benchmark and our methods perform favorably against state-of-the-art algorithms., CVPR2020 camera ready

The inaccuracy of road intersection in digital road map easily brings serious effects on the mobile navigation and other applications. Massive traveling trajectories of thousands of vehicles enable frequent updating of road intersection topology. In this paper, we first expand the road intersection detection issue into a topology calibration problem for road intersection influence zone. Distinct from the existing road intersection update methods, we not only determine the location and coverage of road intersection, but figure out incorrect or missing turning paths within whole influence zone based on unmatched trajectories as compared to the existing map. The important challenges of calibration issue include that trajectories are mixing with exceptional data, and road intersections are of different sizes and shapes, etc. To address above challenges, we propose a three-phase calibration framework, called CITT. It is composed of trajectory quality improving, core zone detection, and topology calibration within road intersection influence zone. From such components it can automatically obtain high quality topology of road intersection influence zone. Extensive experiments compared with the state-of-the-art methods using trajectory data obtained from Didi Chuxing and Chicago campus shuttles demonstrate that CITT method has strong stability and robustness and significantly outperforms the existing methods.

Network Function Virtualization (NFV) enables service flexibility and cost reduction by replacing traditional hardware middle-boxes with Virtual Network Functions (VNFs) running on general-purpose servers. Normally, network traffic usually needs to pass through several VNFs in a particular order. This phenomenon is known as Service Function Chaining (SFC). How to place SFCs with minimal resource is still an open problem. Most of the existing work thinks it's difficult to find the placement solution for all SFCs as a whole, they, instead, consider each demand sequentially, and deploy SFCs one by one. But such serial placement lacks consideration of the interrelations among demands and unable to minimize resource. In this paper, we innovatively propose a parallel deployment scheme based on Deep Reinforcement Learning (DRL). It satisfies demands with minimum resource. We design an overall SFC placement scheme for all demands, and deploy all SFCs simultaneously. We evaluate the proposed algorithms using extensive simulations and prototype experiments. The result demonstrates that our parallel deployment approach minimized resource costs compared with other schemes.

Microarray technology is a popular technique that has been extensively applied in cancer diagnosis. Many studies have used high-dimensional microarray data to identify informative features to classify the types of cancer, yet numerous irrelevant features that exist in microarray data may introduce the noise and decrease classification accuracy. Regularization techniques are common methods for feature selection, which can be used to reduce irrelevant features and avoid overfitting. In recent years, different regularization methods have been proposed. Theoretically, the Lq (0 q (0 q (0

The entrenched instability of solar power output throttles its further integration into power grids worldwide. Thus the precise solar power forecasting (SPF) is helpful for the improvement of power grid stability and better exploitation of clean solar energy. As an important role of ultra-short-term SPF, sky images always contain volatile clouds, which results in tempestuous variation of the output of PV plants. Therefore, an accurate model that can capture the mapping relationship between sky image data and solar irradiance data is significant for fulfilling the ultra-short-term SPF. To fill the gap in the content of this research field, this paper proposes two end to end irradiance mapping models based on deep learning technologies, namely convolutional neural network (CNN) and long short-term memory (LSTM) neural network. Then the mapping performance of the above two mapping models is compared to that of traditional artificial neural network (ANN) based model. In all the aforementioned models, it should be noted that the solar irradiance data output by mapping models is in one-to-one correspondence with the input sky image data in time. The deterministic and probability methods are applied to statistically evaluate the mapping result of CNN and LSTM models. Our case study shows that deep learning architectures, especially the CNN based model, are good at mapping sky images to corresponding surface solar irradiance.

Firstly, a comparison study of varied highway capacity methods and key parameters specified in the guidelines and specifications in China and overseas countries for two lane highway on special grade sections were conducted. The comparison results show that exclusive evaluation indicators and parameters for special grade sections have been provided both in China and abroad; moreover, the analysis method in China is different from that of Germany and Indonesia, but is more similar with that of the Highway Capacity Manual in the US. Secondly, according to the common highway capacity analysis methods in China and the US, the recommended capacity equations and part of key parameters are introduced. Finally, case studies based on real projects are applied for comparison analysis between calculated results by varied capacity analysis methods and on-site field data, the final results illustrate well performance of the recommended methods and parameters.

The traditional static gesture recognition algorithm is easily affected by the complex environment which can cause low recognition rate. Therefore, an improved Faster R-CNN network is adopted to improve the gesture recognition accuracy based on 3D TOF (Time of Flight) camera. The contribution of this paper includes: 1) The group own TOF camera is used for data acquisition of 300 images for each of the five gestures, and the images collected by the TOF camera can reduce the effect of complex environments and provide depth information of gestures; 2) The gesture classification is improved by adding the Inception v3 in the Faster R-CNN. Experiments are carried out to show that the improved network has higher accuracy and stronger robustness, compared with traditional Faster R-CNN, and the localization of gestures in space is accurate with a small number of datasets.

Cancer diseases have serious influence on people's live, but the-state-of-art machine learning approaches have the potential to decrease cancer death rates by formulating prevention strategies for cancer treatment. Some supervised learning methods have been used to give early warning of cancer successfully by using gene expression data, but the most prominent challenge is the insufficient labeled samples of biological data, especially in cancer datasets. It may cause the training model with over-fitting. Therefore, semi-supervised method approaches, such as self-training, are designed to improve the model performance by utilizing the unlabeled samples which are more than the labeled samples in some biological data. The high-quality pseudo-labeled sample is selected to train for improvement of model based curriculum learning paradigm. Of note, the self-training can meet above requirements. Next samples with high quality are pseudo-labeled by self-training. Afterward, put in the training dataset to increase the number of available labeled samples, and the cycle will continue until all unlabeled samples are annotated. The ensemble learning is a widely learning method, which can improve the model classification performance by combining multi-week classifier style. It can help self-training to learn the initial classifiers using the labeled samples. Therefore, in this study an ensemble self-training learning(ESTL) method is proposed to selects the unlabeled samples with high-quality more effectively and improves the robustness of the model. Compared with other tradition classification algorithms, the results of real cancer dataset showed that our proposed ESTL approach improve about 5% in classification performance.

In this paper, by introducing a tuning parameter, an explicit iterative algorithm is constructed for solving the continuous Lyapunov matrix equations associated with Ito stochastic systems. A necessary and sufficient condition is provided to guarantee the convergence of the proposed algorithm by using the Kronecker product. Moreover, some easily verifiable convergence results are developed for the proposed algorithm based on the obtained necessary and sufficient condition. In addition, a method is given for the selection of the optimal tuning parameter. Finally, an example is applied to verify the effectiveness of the presented algorithm.

In recent years, compressed sensing (CS) has been proposed and successfully applied to speed up the acquisition in dynamic MRI. However, how to improve the quality of dynamic MRI is still a worthwhile question. Recently, a low-rank plus sparse (L+S) matrix decomposition model with S 1 and L 1 regularizations is proposed for reconstruction of under-sampled dynamic MRI with separation of background and dynamic components. It can effectively detect dynamic information in the process of imaging. In our work, we propose an improved L+S matrix decomposition model with S 1/2 and L 1/2 regularizations in order to improve the quality of original separation. To solve the model, we use an iterative half-thresholding decomposition algorithm. Finally, empirical results show that the new model can produce better performance and capture more completed dynamic information than the existing model.

Compressed sensing (CS) theory has attracted much attention in the field of signal and image processing. In this paper, the ideas and methods of the image super-resolution (SR) reconstruction combined with CS is studied. The challenge is how to reconstruct a SR image when only one low- resolution (LR) image is available. Regularization methods are the important image SR techniques. Recently, the transform-invariant directional total variation approach with transform-invariant low-rank textures based on Schattenp=1 and L1-norm penalties (TI-DTV+TILT1) has capability of achieving high-quality SR at up-sampling factors. In this paper, we investigate a novel TI-DTV model with TILT based on Schattenp=1/2 (S1/2 -norm) and L1/2-norm penalties (TILT1/2). Moreover, inspired by the alternating direction method of multipliers (ADMM), we propose the alternating threshold iterative algorithm for the new model. Numerous experiments show the proposed method can identity more important information of image, make edges align mostly horizontally and vertically following the low-rank structure and decrease the jagged artifacts along the diagonal line and arcs of high quality SR.

Event stream processing has been used to construct many mission-critical event-driven applications, such as business intelligence applications and collaborative intrusion detection applications. In this paper, we argue that event stream processing is also a good fit for autonomic computing and describe how to design such a system that is resilient to both hardware failures and malicious attacks. Based on a comprehensive threat analysis of event stream processing, we propose a set of lightweight mechanisms that help achieve Byzantine fault tolerant event processing for autonomic computing. The mechanisms consist of voting at the event consumers and an on-demand state synchronization mechanism triggered when an event consumer fails to collect a quorum of matching decision messages. We also introduce an evidence-based safe-guarding mechanism that prevents a faulty event consumer from inducing unnecessary rounds of state synchronization.

Complex event processing has become an important technology for big data and intelligent computing because it facilitates the creation of actionable, situational knowledge from potentially large amount events in soft realtime. Complex event processing can be instrumental for many mission-critical applications, such as business intelligence, algorithmic stock trading, and intrusion detection. Hence, the servers that carry out complex event processing must be made trustworthy. In this paper, we present a threat analysis on complex event processing systems and describe a set of mechanisms that can be used to control various threats. By exploiting the application semantics for typical event processing operations, we are able to design lightweight mechanisms that incur minimum runtime overhead appropriate for soft realtime computing.

The devices working on-line would still have varied of problems. Even they have passed active conformance testing. So an on-line testing approach is needed to make sure the conformance of the devices. In this paper we propose a passive conformance testing method applied to 6LoWPAN neighbor discovery protocol and header compression protocol on-line test. This test method can begin the test at any time while the IUT is working on-line without human intervention. In addition, we add structure inspection algorithms to test the conformance of data structures. Several experiments have been performed to verify the validation of the testing method and the results have proved the test method can be applied to the 6LoWPAN conformance on-line test.

A remote debugging system for OpenMP parallel program is presented in this paper. The system consists of two parts, namely, an integrated debugging environment running on the clent-side and a background daemon running on the server-side. Information exchange between the two sides is accomplished through the socket-based network communication technology. The remote debugging function is realized by the automatic instrumentation technology at code level based on syntax tree. Instance tests show that the system can provide remote users with correctness checking, performance analysis and other debugging functions for OpenMP parallel programs. The user interface is simple and very easy to use.

A new threaded-tree immunization method is proposed in this paper. It refers to searching out highly connected nodes in the network to immunize, by querying a random selected initial node to get its neighbors and then do the same thing to these neighbor nodes recursively. We apply the SIS model on it to do some simulation. And a comparison is then made of the threaded-tree strategy with acquaintance immunization strategy and random immunization strategy, which leads a result that our method is more effective and practical.

The theory of chaos has been widely used for image encryption because of its excellent cryptography characteristics and intrinsic features of image. In this paper, a new image encryption scheme based on chaos is proposed. This new scheme using the principle of magic cube transformation to shuffle the position of image pixels and modify the pixel values with pseudorandom sequences generated by compound chaotic maps. The simulations results show that the image encrypted by this new scheme can not be recognized. Furthermore, the proposed scheme is secure against almost all kinds of attacks.

Three-dimensional diversion structure of open channel is simulated by k - e two equation model and VOF model, the hydraulic characteristics is obtained including flow pattern, flow field distribution, water surface profile and split ratio and so forth. The flow field of diversion region is strong three-dimensional, the relative turbulence intensity increase as the discharge is increased. When the discharge is kept a high value, the flow is narrowed fiercely in the contract section of main channel. The flow engagement style behind splitter pier correlates closely to downstream water stage. It is conclude that the results of numerical is agree well with the results of experimental measured in terms of water surface profile and split ratio by comparison the numerical findings and experiment results.

Intra- and Inter-flow interferences are the principal causes of network utilization degradation in MR-MC (Multi-Radio Multi-Channel) wireless mesh networks. To minimize these interferences, we propose in this paper a channel assignment algorithm and incorporate it in an on-demand routing protocol called the HCARP (Hybrid Channel Assignment Routing Protocol). In addition, HCARP uses load balancing in route discovery to avoid i°hot-spotsi± with severe interference and to find a route satisfying QoS requirements of the incoming flows. Simulation results and test bed experiments demonstrate that intra- and interflow interferences can be mitigated significantly and the overall network throughput is improved considerably. In addition, our algorithm outperforms the one based on superimposed code if the strength of the superimposed code is less than the interferer number of a node. Our simulation results show that HCARP has the capability of circumventing congested areas in the network and enhancing the QoS support for bandwidth-intensive traffic.

An adaptive control scheme based on rigid control for a new model, in which a rigid manipulator handled a flexible payload which is simultaneously driven by an unknown force, is presented. The model of the whole system is derived. The uncertainty of the parameters of the manipulator is considered in the adaptive control scheme. The relationship between two internal forces is analyzed during the establishment of the mathematical model of the whole system. The adaptive rigid controller, which can guarantee the success of trajectory tracking while vibration is suppressed in the absence of the gravity force, is established. The disturbance aroused by part of the unknown force can be avoided using the predetermined compensation. Stability of the system is proven through Lyapunov stability theory. Simulations demonstrate the validity of the proposed controller.

A watershed image segmentation method based on wavelet transform is proposed. Firstly, source image was filtered by multi-scale morphological filtering and the filtered image was decomposed by wavelet, then the low-frequency approximate image was segmented into many small regions by watershed algorithm and these regions were merged according to some region mergence rules to get the original segmentation image. Finally, the original segmentation image was projected into a full-resolution image by inverse-wavelet transform to get the final segmentation image. Utilizing multi-scale morphological filtering and the multi-resolution characteristic of wavelet transform, this method solves effectively the problems of sensitizing to noises and over-segmentation phenomenon of watershed algorithm, and the computing speed is improved.

An image fusion algorithm based on RBF neural networks for multi-focus image is presented in this paper. Four features are defined and proved to indicate the clarity of an image block, and these features are extracted and fed into the neural networks, which then learns to determine which source image is clearer at that particular physical location. The comparison of the algorithm proposed in this paper and the DWT-based one is done. Experimental results show that the performance of the method proposed in this paper is superior to that of DWT-based one.