欢迎访问《图学学报》 分享到:

图学学报

• 图像处理与计算机视觉 • 上一篇    下一篇

基于时域卷积网络的多尺度双线性天气预测模型

  

  1. (1. 南瑞集团有限公司,江苏 南京 211106; 2. 东北电力大学电气工程系,吉林 吉林 132012; 3. 南京大学计算机软件新技术国家重点实验室,江苏 南京 210023)
  • 出版日期:2020-10-31 发布日期:2020-11-05
  • 通讯作者: 路 通(1976),男,江苏盐城人,教授,博士,博士生导师。主要研究方向为视频与图像数据处理、模式识别与计算机图形学等。 E-mail:lutong@nju.edu.cn
  • 作者简介:孔 震(1979?),男,江苏仪征人,高级工程师,硕士。主要研究方向为大数据、信息通信、人工智能及其应用等。 E-mail:kongzhen@sgepri.sgcc.com.cn
  • 基金资助:
    国家电网有限公司总部科技项目

Multi-scale and bilinear models based on temporal convolutional network

  1. (1. Nari Group Corporation, Nanjing Jiangsu 211106, China; 2. Department of Electrical Engineering, Northeast Electric Power University, Jilin Jilin 132012, China; 3. National Key Lab for Novel Software Technology, Nanjing University, Nanjing Jiangsu 210023, China)
  • Online:2020-10-31 Published:2020-11-05
  • Contact: LU Tong (1976-), male, professor, Ph.D. His main research interests cover video and image data processing, pattern recognition and computer graphics. E-mail:lutong@nju.edu.cn
  • About author:KONG Zhen (1979-), male, senior engineer, master. His main research interests cover information and communication, artificial intelligence and applications. E-mail:kongzhen@sgepri.sgcc.com.cn
  • Supported by:
    Scientific Foundation of State Grid Corporation of China

摘要: 针对时域卷积网络(TCN)提取能力受卷积层感受野限制,难以对天气数据中的季节 性信息和长时信息进行有效提取与分析的问题,提出了一个新的基于 TCN 的多尺度双线性天气 预测模型。该模型由 TCN 层和双线性汇合层 2 部分组成,时域卷积层包含双路 TCN,每个 TCN 利用历史观测数据独立提取特征,除卷积核尺度之外,其他网络参数均保持一致。多尺度的 网络组合可以更深入挖掘数据中潜在关联信息;时域卷积层的输出作为双线性汇合层的输入 进行双线性融合,规范化后得到最终输出,即对未来天气的预测值,进一步提升模型的特征 表示能力。在公开的天气预测数据集上与 5 个基准方法进行对比,实验结果表明所提方法的预 测结果准确率更高;此外,对比 TCN,多尺度双线性天气预测模型面对长时数据信息时表现更 加稳定。

关键词: 时域卷积网络, 多尺度融合, 双线性汇合, 数据挖掘, 天气预测

Abstract: The learning capability of such convolutional network models as temporal convolutional network (TCN) is limited to the scale of receptive field, and sometimes it is difficult to mine data from seasonal or long-term information, such as weather data for deep convolutional networks. In order to address this problem, a novel multi-scale and bilinear convolutional neural network was proposed, which was composed of the TCNs layer and bilinear layer. The TCNs layer contained two temporal convolutional networks, each of which extracted useful features independently from the historical data. All TCNs had the same parameters expect for the kernel size, thereby further mining the underlying relevant information in data. The output of the TCNs layer was employed as the input of the bilinear layer, and after normalization, the final output could be eventually calculated, which enhanced feature representation capabilities of the model. The comparisons with five methods on the public datasets of weather forecast show that the proposed method can effectively improve the accuracy and performance in terms of the long-time data much better than the TCN.

Key words: temporal convolutional network, multi-scale convolutional, bilinear fusion, data mining; weather forecast