Results 1 to 10 of about 2,013 (147)
Design of Sequential Wakeup Compute-In-Memory Controller Based on Convolutional Neural Network [PDF]
With the development of artificial intelligence, the demand for intelligent image processing on edge devices has significantly increased. At present, edge devices mainly face issues such as limited energy and low throughput.
SONG Qingzeng, LIU Xiangdong, XU Kangwei, LIU Jiahui, REN Erxiang, LUO Li, WEI Qi, QIAO Fei
core +1 more source
基于双重注意力变换模型的分布式屋顶光伏变电站级日前功率预测 [PDF]
全球能源互联网分布式屋顶光伏地理位置分散，受地理环境遮挡和多种气象因素影响，导致光伏出力特性存在差异，给变电站级分布式屋顶光伏日前功率预测造成挑战。针对上述问题，提出了一种基于双重注意力变换模型的分布式屋顶光伏变电站级日前功率预测方法。首先，基于动态时间规整算法计算分布式光伏用户出力特性间的相似度，并基于凝聚层次聚类法将其划分成若干类；然后，利用自主注意力网络学习各时间步间的时序关联特性，通道卷积注意力机制学习多特征变量间的相关性，构建日前功率预测模型；最后，将每一类日前预测结果相加，实现变电站级日前功率预测 ...
王光华 +5 more
doaj +1 more source
Zonal prediction of the heights of water-conducting fracture zones under varying overburden types in North China-type coalfields [PDF]
BackgroundThe gradual increase in the exploitation depth and intensity of coal resources in the North China-type coalfields has caused problems such as overburden movement and damage, as well as fracture evolution.
Dongjing XU +6 more
core +1 more source
Fault diagnosis of marine electric thruster gearbox based on MPDCNN under strong noisy environments [PDF]
ObjectivesTo address the performance degradation in fault diagnosis of rotating machinery caused by noise interference in practical applications, a novel fault diagnosis approach based on Mel-frequency cepstral coefficients (MFCC) and a parallel dual ...
Qianming SHANG +4 more
core +1 more source
Gaoya dianqi, 2017 对局部放电进行有效识别可以为评估变压器设备绝缘状况提供科学的参考依据,然而局部放电类型的识别往往需要人为地提取描述特征,适应性很差。针对此问题,提出一种基于卷积神经网络的智能识别新方法。根据视觉注意机制分割出放电信号图像,并将灰度化和双线性插值归一化处理的图像作为卷积神经网络的输入。该方法模拟人脑的机制来解释数据,可以直接对采集到的放电信号图像进行自动特征学习与模式识别。实验中对4种典型放电类型的识别率超过了94%,显著优于传统的方法。试验结果表明,该方法无需进行复杂的特征提取 ...
刘兵, 郑剑
doaj
, 2018 深度神经网络在人工智能的应用中,包括计算机视觉、语音识别、自然语言处理方面,取得了巨大成功.但这些深度神经网络需要巨大的计算开销和内存存储,阻碍了在资源有限环境下的使用,如移动或嵌入式设备端.为解决此问题,在近年来产生大量关于深度神经网络压缩与加速的研究工作.对现有代表性的深度神经网络压缩与加速方法进行回顾与总结,这些方法包括了参数剪枝、参数共享、低秩分解、紧性滤波设计及知识蒸馏.具体地,将概述一些经典深度神经网络模型,详细描述深度神经网络压缩与加速方法,并强调这些方法的特性及优缺点.此外 ...
吴永坚 +4 more
core
Dynamical class-weighted-based convolutional neural networks attack detection model [PDF]
The intrusion detection system (IDS) as a core component of IoT security defense, was directly impacted in its performance, which in turn affected the overall security of the network.
FAN Rong
core +1 more source
Gaoya dianqi为了提高电力系统暂态电压在系统拓扑结构发生变化时能够稳定评估，以及提高在时空方面的特征提取能力，提出一种图卷积网络与循环神经网络相融合的方法。首先，引入图卷积网络对电力数据进行图表示，将电力系统建模为网络结构，自动学习电压节点的特征表示。其次，提出使用循环神经网络来处理暂态电压数据的时间依赖关系，捕捉暂态电压数据的时序特征。然后，提出自适应增强模块，用于将两个输出特征表示相互融合，提高模型在系统拓扑结构上的时空特征提取能力。最后，通过算例验证表明，相比于传统的评估模型，所提方法具有更高的预测精度和有效性。
徐焕 +5 more
doaj
Bi-directional Pre-trained Network for Single-station Seismic Waveform Analysis [PDF]
The application of machine learning, particularly deep learning methods, is becoming increasingly widespread in seismology, achieving near-human accuracy in tasks such as phase detection and event classification.
Lu LI +4 more
core +1 more source
Dianci bileiqi, 2020 针对无人机巡检输电线路图像中存在绝缘子串定位难点,笔者在分析散射变换原理和卷积神经网络（CNN）的基础上,通过低通滤波器作散射系数处理,结合Gram矩阵法来降低绝缘子串背景信息的噪声干扰,以增强低频系数的边缘纹理特征,结合SSD网络框架实现了CNN对绝缘子串实时定位计算的高效性。实验结果表明:该方法在保证实时计算的前提下,与传统SSD网络框架相比,召回率和交并比分别提升了1. 04%和1. 38%。
潘翀 +5 more
doaj