Search

Transformers are important equipment in the power system. Aiming at the collection, transmission, storage, and processing of massive high-dimensional vibration data in the process of transformer vibration online monitoring, this paper proposes a transformer vibration signal feature extraction method based on compressed sensing and wavelet packets. Firstly, the PartHadamard measurement matrix is used to compress the transformer vibration signal, and then the characteristics of the transformer vibration signal are extracted based on the wavelet packet decomposition and normalized wavelet information entropy. The proposed method completes feature extraction on the measured vibration data of the transformer, and three classification algorithms are applied to carry out simulation experiments. The results show that the proposed feature extraction method can effectively extract the vibration characteristics reflecting the operating state of the transformer while greatly reducing the dimension of vibration data, which provides a reference for the online monitoring of transformer vibration.

Since the Generative Adversarial Networks (GANs) was proposed, researches on image generation attract many scholars’ general attention and good graces. Traditional GANs generate a sample by playing a minimax game between generator and discriminator. In this paper, we propose a new method called EmotionGAN for generating facial expression. Specifically, the inverse of the generator is firstly utilized to establish the mapping between the input and feature vector. Then the Generalized Linear Model (GLM) is used to fit the changing direction of different expressions in the feature space, which provide a linear guidance to the feature vector along the expression axis, and thus spatial distribution consistence with the target feature vector is assured. Finally the generator is applied to reconstruct the facial image of the expression. By controlling the intensity of the feature vector, the generated image can be smoothly changed on a specific expression. Experiments have shown that EmotionGAN can quickly generate face images with arbitrary expressions while ensuring identity information is not changed, and the image attributes are more accurate and the resolution is higher.

Person re-identification is a difficult topic in computer vision. Some study think that current deep learning methods is biased to capture the most discriminative features and ignore low-level details, more serious is it pay too much attention on relevance between background appearances of person images. It might limit their accuracy or makes them needlessly expensive for a not best performance. In this paper, we carefully design the Weak Reverse attention with Context Aware Network (WRCANet). Specifically, by merging weak reverse attention network and content aware module, the model can not only remove the background noise to extract the main information of persons, but also suppress the loss of local detailed information as the network deepens. We experiment on the Market-1501, DukeMTMC-reID and CUHK03, and the results show that our method achieves the state-of-the-art performance.

This paper proposes a simple and efficient regularization method, called PickPatch, for face recognition based on a deep convolutional neural network (DCNN). The proposed method randomly selects patches in the input face image and the intermediate feature maps as the activation region according to facial landmarks during the training phase. PickPatch is an approximation method that trains a series of models for different face patches and provides a combined model. This strategy introduces the idea of model combination for multiple face patches but does not change the model structure, which is both simple and efficient. Experiments on the public LFW database demonstrate that the proposed regularization method based on current deep convolutional neural networks can achieve obvious improvements of face recognition accuracy.

With the improvement of power conservation awareness and the development of the power market, load forecasting is playing an important role gradually. Effective load forecasting can help power system operators to develop appropriate scheduling strategies and help users plan their power consumption rationally. Considering the important position of load forecasting in the future power system field, this paper focus on this field. Based on the measured data from Australia, this paper considers the environmental and social factors affecting the power consumption in the region. The LSSVM algorithm is used for short-term load forecasting. Firstly, the invalid data is eliminated by data preprocessing, the missing data is completed, and the data of the content is included. The normalization process is carried out, and then the processed data is used for prediction. Finally, the effectiveness of the Least squares support vector machine algorithm load prediction after considering various factors is verified by comparison with the traditional neural network algorithm.

This paper presents a self-sustaining solar energy harvester (EH) based on 130nm process, which consists of a low voltage cold start (CS) circuit and a high efficiency boost charger. The CS circuit makes the boost charger self-sustaining and start the boost circuit from the 100mV input. An adaptive switch frequency modulation (SFM) and switch capacitor modulation (SCM) are proposed to minimize the switch loss and conduction loss of boost charger, thus reduce overall circuit power consumption. The proposed boost charger utilizes a dual voltage technique for reducing the power dissipation of the internal circuit. The efficiency of the boost charger ranges from 43% at V IN = 100mV to 65% at V IN = 300mV.