Search

The frequency security problem of power system is highlighted as wind power penetration increases yearly, the eigenvalue analysis method based on the deterministic model is difficult to accurately evaluate the small signal stability of power system. The Weibull probability distribution is used to describe the uncertainty of wind speed. In this paper, the shape and scale parameters of the Weibull distribution are calculated by using the maximum likelihood approach and combining with the measured data, and the stability of small signal with large-scale wind power access is analyzed by the probability distribution method. Meanwhile, considering that the lack of inertia of the system caused by high-penetration wind power access can easily lead to the problem of frequency stability, it is proposed to take the maximum damping ratio of regional oscillation mode as optimization objective, virtual inertia and pitch angle control parameters as variables, and frequency stability as constraints, an small signal stability optimization model of wind power access to power system is established according to the proposed probabilistic small signal stability model. The sensitivity algorithm is used to obtain the optimal solution for the virtual inertia and pitch angle control parameters, and the validity of the proposed method and model is verified by two-area four-machine examples. Simulation results show that by optimizing the virtual inertia and pitch angle parameters of wind turbines, the small signal stability of the high-penetration wind power access system can be improved under the condition of ensuring sufficient inertia of the system, which provides a certain theoretical basis for the safe and reliable operation of wind power large-scale cluster access power system.

Spontaneous speech emotion recognition is a new and challenging research topic. In this paper, we propose a new method of spontaneous speech emotion recognition on the basis of binaural representations and deep convolutional neural networks (CNNs). The proposed method initially employs multiple CNNs to learn deep segment-level binaural representations such as Left-Right and Mid-Side pairs from the extracted image-like Mel-spectrograms. These CNNs are fine-tuned on target emotional speech datasets from a pre-trained image CNN model. Then, a new feature pooling strategy, called block-based temporal feature pooling, is proposed to aggregate the learned segment-level features for producing fixed-length utterance-level features. Based on the utterance-level features, linear support vector machines (SVM) is adopted for emotion classification. Finally, a two-stage score-level fusion strategy is used to integrate the obtained results from Left-Right and Mid-Side pairs. Extensive experiments on two challenging spontaneous emotional speech datasets, including the AFEW5.0 and BAUM-1s databases, demonstrate the effectiveness of our proposed method.

The minimum neighborhood and combinatorial property are two important indicators of fault tolerance of a multiprocessor system. Given a graph G, θG(q) is the minimum number of vertices adjacent to a set of q vertices of G (1 ≤ q ≤ |V(G)|). It is meant to determine θG(q), the minimum neighborhood problem (MNP). In this paper, we obtain θAG0(q) for an independent set with size q in an n-dimensional alternating group graph AGn, a well-known interconnection network for multiprocessor systems. We first propose some combinatorial properties of AGn. Then, we study the MNP for an independent set of two vertices and obtain that θAGn(2) = 4n - 10. Next, we prove that θAGn(3) = 6n -16. Finally, we propose that θAGn(4) = 8n - 24.

One key challenging issues of facial expression recognition (FER) in video sequences is to extract discriminative spatiotemporal video features from facial expression images in video sequences. In this paper, we propose a new method of FER in video sequences via a hybrid deep learning model. The proposed method first employs two individual deep convolutional neural networks (CNNs), including a spatial CNN processing static facial images and a temporal CN network processing optical flow images, to separately learn high-level spatial and temporal features on the divided video segments. These two CNNs are fine-tuned on target video facial expression datasets from a pre-trained CNN model. Then, the obtained segment-level spatial and temporal features are integrated into a deep fusion network built with a deep belief network (DBN) model. This deep fusion network is used to jointly learn discriminative spatiotemporal features. Finally, an average pooling is performed on the learned DBN segment-level features in a video sequence, to produce a fixed-length global video feature representation. Based on the global video feature representations, a linear support vector machine (SVM) is employed for facial expression classification tasks. The extensive experiments on three public video-based facial expression datasets, i.e., BAUM-1s, RML, and MMI, show the effectiveness of our proposed method, outperforming the state-of-the-arts.

The failure of some key vertices in a network, due to either attacks or malfunctioning, may break the network into several disconnected components-the vertices within each component are connected, but there are no connections between components. When this happens, two measures should be of concern: (1) the number of components in the remaining network; and (2) the size of each component. When a given number of “break vertices”are removed from a network, we hope to have as few components as possible in the remaining network. On the other hand, it is certainly desirable that the components are as large as possible. Therefore both the number of components and the size of the maximal component after removing certain vertices can be a metric of a network's fault tolerability, an important dimension of its robustness. In this paper, we examine the split-star, denoted S2n, in terms of this metric. We prove that when an arbitrary subset of vertices D with |D| ≤ 6n-15 is removed from S2n, the remaining network will have at most 3 components, with the largest component having at least n! - |D| -3 vertices. With |D| ≤ 8n -21, the remaining network has at most 4 components, with the largest component's size at least n!-|D|-5. As a theoretical application, the r-component connectivity of S2n is estimated by using the obtained maximal component and the minimal neighbor-set of independent set of size r. Moreover, we present a bound of r-component edge connectivity on S2n by considering the minimal neighbor edge-set of appropriate substructures in S2n.