检索结果-内蒙古大学图书馆

Multivariable passive method for detection of islanding events in renewable energy based power grids

IET RENEWABLE POWER GENERATION 2022年第3期16卷 497-516页

作者： Swarnkar, Nagendra Kumar Mahela, Om Prakash Khan, Baseem Lalwani, Mahendra Rajasthan Tech Univ Dept Elect Engn Kota 324010 India Rajasthan Rajya Vidyut Ptasatan Nigam Ltd Power Syst Planning Div Jaipur Rajasthan India Hawassa Univ Dept Elect & Comp Engn Hawassa 05 Ethiopia

Penetration of distributed generation resources (DGR) in power grid is rapidly increasing to meet future energy demand efficiently. It helps in mitigating the problems of high carbon emission, green house effect, and increased cost of oil and natural gases. Island formation of a part of the renewable energy based power grid (REPG) creates hazards to operating personnel and equipments. Hence, efficient identification of islanding condition and disconnection of DGR immediately become essential for avoiding damage to equipment, personnel safety hazards and grid protection. This paper investigated a signal processing based passive islanding detection method (IDM) which is effective with the availability of multi-DGR. A islanding detection factor (MIDF) using multi-variables is proposed which is computed using features extracted by processing the voltage signal, negative sequence current (NSC) and negative sequence voltage (NSV) using Stockwell transform (ST) and Hilbert transform (HT). Total harmonic distortions of voltage (THDv) and current (THDi) are also considered to compute MIDF. Islanding event is recognized and differentiated from non-islanding conditions by comparing the MIDF with two threshold values. This method effectively detects the islanding event and discriminates it from the operational and faulty events. IDM has small non-detection zone (NDZ) and low computational time. Performance of IDM is not affected by noisy environment. Effectiveness of the IDM is established by testing for different event incidence angles, various fault impedances, different values of loads and capacitors, and different penetration levels of RE. Efficacy of IDM is established by testing the IDM through an elaborated study on IEEE-13 bus distribution grid interfaced with wind power plant (WPP) and solar power plant (SPP). Further, IDM is effectively implemented to detect the islanding event created on a practical distribution network. Effectiveness of IDM is better in comparison t

关键词： passive islanding detection method DGR penetration grid protection greenhouse effect power engineering computing distributed power generation reactive power solar power stations THDi natural gases RPNN total harmonic distortion of voltage THDv Hilbert transform islanding event detection neural nets discrete wavelet transform MIDF noisy environment personnel safety hazards islanding detection factor multiDGR artificial neural network discrete wavelet transforms Slantlet transform distributed generation resources nondetection zone event incidence angles wind power plant power grids signal detection empirical mode decomposition faulty events high carbon emission hazards islanding condition signal processing voltage signal processing Ridgelet probabilistic neural network feature extraction renewable energy based power grid green house effect IEEE-13 bus distribution grid negative sequence voltage island formation probability harmonic distortion future energy demand operational events total harmonic distortion current operating personnel NSC Hilbert transforms EMD IDM solar power plant nonislanding conditions ANN multivariable passive method REPG WPP DWT voltage angle active power wind power plants Stockwell transform

来源：评论

学校读者我要写书评

暂无评论

Direct image classification using an on-chip photonic deep neural network 27

Direct image classification using an on-chip photonic deep n...

引用

Conference on Integrated Optics - Devices, Materials, and Technologies XXVII

作者： Ashtiani, Farshid Geers, Alexander J. Aflatouni, Firooz Univ Penn Dept Elect & Syst Engn 200 S 33rd St Philadelphia PA 19104 USA Nokia Bell Labs 600 Mt Ave Murray Hill NJ 07974 USA

ISBN: (纸本)9781510659537;9781510659544

Artificial neural networks have been a major driving force in a wide variety of technologies from pattern recognition and natural language reprocessing to medical diagnosis. neural networks can be trained to perform a certain set of tasks and are typically formed by the interconnection of multiple layers of neurons to perform linear and nonlinear computations. Highly reconfigurable clock-based digital signal processors have been the main hardware platform to implement such networks. However, clock frequency and memory access time are two main contributors in limiting the processing speed. Integrated photonics benefits from the large available bandwidth at optical frequencies and low-loss interconnects, and has proved to be a promising candidate to address the challenges of electronic systems. Most of the demonstrated photonic neural networks have been either using bench-top setups, or the partial integration of some of the linear and/or nonlinear computation blocks. A scalable solution that can integrate all computation blocks has been an open problem. Here, we demonstrate the first end-to-end photonic deep neural network (PDNN) for direct image classification. On-chip image formation of the target object using a 2D array of grating couplers and direct processing of the received optical signals through linear (optical) and nonlinear (opto-electronic) computation, enable an end-to-end classification time of 570 ps. Using a uniformly distributed supply light enables scaling the PDNN to networks with more layers of neurons. The PDNN chip is used to demonstrate classification of 2- and 4-class datasets of handwritten letters and the system achieves accuracies higher than 93.8% and 89.8%, respectively.

关键词： Photonic neural network Silicon photonics Integrated photonics Image classification

来源：评论

学校读者我要写书评

暂无评论

Iteration number-based hierarchical gradient aggregation for distributed deep learning

引用

JOURNAL OF SUPERCOMPUTING 2022年第4期78卷 5565-5587页

作者： Xiao, Danyang Li, Xinxin Zhou, Jieying Du, Yunfei Wu, Weigang Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou Peoples R China Natl Supercomp Ctr Guangzhou Peoples R China

distributed deep learning can effectively accelerate neural model training, which employs multiple workers at a cluster of nodes to train a neural network in a parallel way. In this paper, we propose InHAD, an asynchronous distributed deep learning protocol, whose key novelty lies in the design of hierarchical gradient communication and aggregation. The local aggregation is conducted inside a computing node to aggregate the gradients produced by local workers, while global aggregation is conducted at the parameter server to calculate new model parameters based on results of local aggregations. An iteration number-based mechanism is designed to guarantee the convergence of the training. Worker nodes keep sending IN to parameter server and the latter decides when to pull gradients from workers by counting IN and updates global model weights/parameters. With the IN-based hierarchical aggregation technique, InHAD can save communication cost by reducing the number of gradients transferred and speed up the convergence by limiting the staleness of gradients. We conduct extensive experiments at the Tianhe-2 supercomputer system to evaluate the performance of InHAD. Two neural networks are trained on two classical datasets, and similar protocols like Horovod and ASP are tested for comparisons. The results show that InHAD can achieve much higher acceleration than ASP and nearly the same accuracy as Horovod .

关键词： distributed deep learning Gradient aggregation Parallel processing Parameter server Asynchronous training

来源：评论

学校读者我要写书评

暂无评论

Construction of Computer Vision Guidance System for Spacecraft Orbit Based on Artificial Intelligence Algorithm

Construction of Computer Vision Guidance System for Spacecra...

引用

2023 International Conference on Internet of Things, Robotics and distributed Computing, ICIRDC 2023

作者： Qiu, Rui Ma, Guangcheng Fu, Hao Harbin Institute of Technology Heilongjiang Harbin150006 China Beijing Institute of Spacecraft Environment Engineering Beijing100094 China

ISBN: (纸本)9798350375275

Accurate navigation and positioning of spacecraft is the key factor to ensure the successful implementation of its mission. In order to cope with the complex and changeable space environment, a computer vision guidance system for spacecraft orbit based on artificial intelligence (AI) algorithm is proposed in this study. Based on the deep learning algorithm, the system aims to realize an efficient and robust navigation system by processing and analyzing the visual information of spacecraft orbit in real time. In this study, neural network can approximate any nonlinear function, the relative motion state of the spacecraft at the current moment and the remaining time of the pursuit game are taken as inputs, and the maneuvering acceleration of the spacecraft during the pursuit process is taken as output. The internal relationship contained in these data is studied by using deep neural network, and an online generation method of maneuvering strategy of the spacecraft pursuit game with fixed stay based on deep neural network is obtained. By training on large-scale data sets, we demonstrate the generalization ability of the deep learning model in unknown orbit scenes. The good generalization of the model makes the system have the ability to adapt to different tasks and environments, which provides strong support for the reliable implementation of practical tasks. © 2023 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Privacy-preserving short-term load forecasting with convolutional neural networks using secure computation 9

Privacy-preserving short-term load forecasting with convolut...

引用

9th International Symposium on Sensors, Mechatronics, and Automation System, ISSMAS 2023

作者： Yan, Chenyang Yan, Tingting Gao, Ziting State Grid Information & Telecommunication Group Co. Ltd. Beijing100053 China

ISBN: (纸本)9781510672802

This paper aims to explore the use of privacy-preserving computation and convolutional neural networks for short-term electricity load forecasting, with the goal of addressing the challenges of privacy leakage in data sharing. To achieve this, techniques such as homomorphic encryption are adopted to encrypt historical load data, protecting the privacy of the electricity load data, while the convolutional neural network is utilized to extract the data's characteristics and predict future electricity loads through a CNN model. This approach ensures the privacy and security of data while simultaneously improving forecasting accuracy. In this study, electricity load data from a city in Southwest China is used for experimentation. The results show that our proposed method can achieve high prediction accuracy (up to 95%) while preserving data privacy. Therefore, the results of this research are of significant importance to the reliability of the power system and privacy protection research. © 2024 SPIE.

关键词： Data modeling Data privacy Convolutional neural networks distributed computing Feature extraction Computer security Deep learning Education and training network security Data processing

来源：评论

学校读者我要写书评

暂无评论

Harnessing Manycore Processors with distributed Memory for Accelerated Training of Sparse and Recurrent Models 38

Harnessing Manycore Processors with Distributed Memory for A...

引用

38th AAAI Conference on Artificial Intelligence (AAAI) / 36th Conference on Innovative Applications of Artificial Intelligence / 14th Symposium on Educational Advances in Artificial Intelligence

作者： Finkbeiner, Jan Gmeinder, Thomas Pupilli, Mark Titterton, Alexander Neftci, Emre Res Ctr Juelich Julich Germany Rhein Westfal TH Aachen Aachen Germany Graphcore Bristol England

ISBN: (纸本)1577358872

Current AI training infrastructure is dominated by single instruction multiple data (SIMD) and systolic array architectures, such as Graphics processing Units (GPUs) and Tensor processing Units (TPUs), that excel at accelerating parallel workloads and dense vector matrix multiplications. Potentially more efficient neural network models utilizing sparsity and recurrence cannot leverage the full power of SIMD processor and are thus at a severe disadvantage compared to today's prominent parallel architectures like Transformers and CNNs, thereby hindering the path towards more sustainable AI. To overcome this limitation, we explore sparse and recur-rent model training on a massively parallel multiple instruction multiple data (MIMD) architecture with distributed local memory. We implement a training routine based on backpropagation through time (BPTT) for the brain-inspired class of Spiking neural networks (SNNs) that feature binary sparse activations. We observe a massive advantage in using sparse activation tensors with a MIMD processor, the Intelligence processing Unit (IPU) compared to GPUs. On training work-loads, our results demonstrate 5-10x throughput gains compared to A100 GPUs and up to 38x gains for higher levels of activation sparsity, without a significant slowdown in training convergence or reduction in final model performance. Furthermore, our results show highly promising trends for both single and multi IPU configurations as we scale up to larger model sizes. Our work paves the way towards more efficient, non-standard models via AI training hardware beyond GPUs, and competitive large scale SNN models.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

A Feature Selection Method Based On Software Defined networks 30

A Feature Selection Method Based On Software Defined Network...

引用

30th IEEE Signal processing and Communications Applications Conference (SIU)

作者： Undar, Desdina Kof Pehlivanli, Zeynep Gul Arsal, Ali ULAK Haberlesme AS Sistem Muhendisligi Ankara Turkey

ISBN: (纸本)9781665450928

Software-defined networking is in the midst of information security challenges. In the case that distributed denial of service attacks occur against a software-defined network controller, network traffic data becomes vulnerable because of the overload. In this study, distributed denial of service attacks in software defined network are detected by using machine learning based models with different datasets. First, certain features are obtained on the software-defined network for the dataset under normal conditions and under attack traffic. Subsequently, a new data set was generated by using the proposed hybrid feature selection method on the existing data set. The proposed feature selection method algorithm is trained and tested with logistic regression, artificial neural network, k-nearest neighbors and Naive Bayes models. The results show that the use of the hybrid method has positive impacts on the performance metrics and provides lower processing time.

关键词： software defined network feature selection machine learning logistic regression artificial neural network k-nearest neighbors naive bayes

来源：评论

学校读者我要写书评

暂无评论

Improved Hypernymy Detection Algorithm Based on Heterogeneous Graph neural network

引用

INTERNATIONAL JOURNAL OF COMPUTATIONAL INTELLIGENCE SYSTEMS 2025年第1期18卷 1-27页

作者： Ren, Li Huang, Jing Jia, Hai-Tao Sun, Shu-Bao Wang, Kai-Shi Wu, Yi-Le Univ Elect Sci & Technol China Univ Elect Sci & Technol Lib Chengdu 611731 Sichuan Peoples R China Univ Elect Sci & Technol China Sch Resources & Environm Chengdu 611731 Sichuan Peoples R China Taiyuan North Inst Automat Control Technol Taiyuan 030900 Shanxi Peoples R China

Concept mapping is a knowledge representation method used to represent and understand concepts, entities, and the relationships between them, which are referred to as hyponymy-hypernymy semantic relations. These relations are primarily used to describe the hierarchical and categorical relationships between different concepts or entities. The detection of hyponymy-hypernymy semantic relations is an important task in the field of natural language processing, crucial to many downstream tasks such as information extraction, automatic reasoning, and personalized recommendations. These tasks often require understanding the semantic relations between concepts or entities in text for more accurate analysis and reasoning. Currently, algorithms for identifying and detecting hyponymy-hypernymy semantic relations face two main challenges: first, candidate hyponymy-hypernymy relation tuples do not exist in the same contextual sentence, failing to meet the co-occurrence requirement;second, distributed algorithms have issues with lexical memory. To address these issues, this paper proposes an improved algorithm for detecting hyponymy-hypernymy relations based on heterogeneous graph neural networks, aiming to detect hyponymy-hypernymy relations in various candidate word sets and then construct a hierarchical system. To meet the co-occurrence requirement of candidate word pairs, an open-source large model is utilized to generate contextual sentences for the candidate pairs. Sub-word features are adopted to capture the intrinsic semantic connections between nested phrases, thus alleviating the issue of reversed predictions for nested phrases. The representation of relation nodes is modified by encoding relation definitions through a pre-trained model, enabling the model to understand the semantic relations between concept nodes. To address the problem of overfitting in traditional graph attention networks, the calculation order of adjacency node aggregation is changed in the heteroge

关键词： Heterogeneous graph neural network Hypernymy detection Hierarchical structure construction Semantic relation understanding

来源：评论

学校读者我要写书评

暂无评论

Research on Ubiquitous Power Internet of Things in Distribution networks Using Adaptive Gradient Algorithm with Convolution neural network 2

Research on Ubiquitous Power Internet of Things in Distribut...

引用

2nd IEEE International Conference on Data Science and network Security, ICDSNS 2024

作者： Liu, Songmei Wanjiang University of Technology Ma'anshan China

ISBN: (纸本)9798350373110

In China, several distribution automation devices are established in distribution network to increase reliability of power supply. The major challenge of ubiquitous electric Internet of Things (IoT) is a resource aspect which enables distributed computational sources. Therefore, the Adaptive Gradient Algorithm with Convolution neural network (Adagrad-CNN) is proposed for ubiquitous electric IoT in distribution networks. The Adagrad is helpful when dealing with sparse features or data that has extensive-range of values. The CNN has multiple-layer architecture for classifying image pixels patterns through processing and it filtered noise captures significant predictive properties. The Adagrad-CNN leads to better generalization and enhance the CNN performance through mitigating issues based on exploding gradients that is essential when dealing with difficult and large datasets in ubiquitous electric IoT. The dataset is collected from various devices and sensors in the power grid and it is preprocessed by Z-score normalization. This preprocessing enhances convergence rates and reduces the large scales. It is essential when features are in different ranges and it prevents particular feature dominance. The Adagrad-CNN attains 96.54%, 94.21 %, 95.73% and 94.59% of accuracy, recall, precision and f1-score. © 2024 IEEE.

关键词： Power distribution networks

来源：评论

学校读者我要写书评

暂无评论

Task-based Acceleration of Bidirectional Recurrent neural networks on Multi-core Architectures 36

Task-based Acceleration of Bidirectional Recurrent Neural Ne...

引用

36th IEEE International Parallel and distributed processing Symposium (IEEE IPDPS)

作者： Sharma, Robin Kumar Casas, Marc Univ Politecn Catalunya UPC Barcelona Supercomp Ctr BSC Comp Sci Dept Barcelona Spain

ISBN: (纸本)9781665481069

This paper proposes a novel parallel execution model for Bidirectional Recurrent neural networks (BRNNs), BPar (Bidirectional-Parallelization), which exploits data and control dependencies for forward and reverse input computations. B-Par divides BRNN workloads across different parallel tasks by defining input and output dependencies for each RNN cell in both forward and reverse orders. B-Par does not require perlayer barriers to synchronize the parallel execution of BRNNs. We evaluate B-Par considering the TIDIGITS speech database and the Wikipedia data-set. Our experiments indicate that BPar outperforms the state-of-the-art deep learning frameworks TensorFlow-Keras and Pytorch by achieving up to 2.34x and 9.16x speed-ups, respectively, on modern multi-core CPU architectures while preserving accuracy. Moreover, we analyze in detail aspects like task granularity, locality, or parallel efficiency to illustrate the benefits of B-Par.

关键词： Deep neural network (DNN) Bidirectional recurrent neural networks (BRNNs) Long-short term memory (LSTM) Gated Recurrent Units (GRU) Task Parallelism

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：