基于 SA-LSTM的电力光缆光功率预测方法
CSTR:
作者:
作者单位:

(1.国网福建省电力有限公司,福州 350001;2.国网福建省电力有限公司电力科学研究院,福州 350007; 3.国网福建省电力有限公司厦门供电公司,福建厦门 361004;4.华北电力大学电子与通信工程系,河北保定 071003)

作者简介:

陈锦山(1986-),男,福建省漳平市人,高级工程师,主要研究方向为电力通信及自动化;王新澜(1997-),女,河北省张家口市人,助理工程师,主要研究方向为电力系统自动化及通信。

通讯作者:

中图分类号:

TN913

基金项目:


Simulated Annealing-Long Short-Term Memory Based Light-Power Prediction Method for Power Cables
Author:
Affiliation:

(1. State Grid Fujian Electric Power Co.,Ltd.,Fuzhou 350001,CHN;2. State Grid Fujian Electric Power Co.,Ltd. Electric Power Research Institute,Fuzhou 350007,CHN;3. State Grid Fujian Electric Power Co.,Ltd. Xiamen Power Supply Company,Xiamen 361004,CHN;4. Department of Electronic and Communication Engineering,North China Electric Power University,Baoding 071003,CHN)

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为提高基于光功率预知光缆异常的准确性,提出一种基于模拟退火长短期记忆(Simulated Annealing-Long Short-Term Memory,SA-LSTM)算法的光功率波动预测方法。首先,对原始光功率数据进行归一化处理;随后,利用模拟退火算法对长短期记忆模型的关键超参数(包括迭代次数、隐藏层单元数和学习率)在大范围内进行全局寻优,以获得最优参数组合;最后,基于优化后的参数建立 LSTM模型,用于预测光功率的未来趋势。实验结果表明,采用最优参数时,所提方法的光功率预测均方根误差为 2. 83×10.3 dB,该误差仅为光功率均值的 0. 14%。相较于自回归移动平均径向基函数算法、差分自回归移动平均支持向量机算法、自回归移动平均长短期记忆算法和差分自回归移动平均长短期记忆算法,预测精度分别提升了 91. 7%、91. 0%、96. 4%和 96. 3%。因此,所提方法能够更加准确地预测电力光缆的光功率变化趋势,具有重要的工程应用价值。

    Abstract:

    To improve the accuracy of predicting anomalies in optical cable systems based on optical power data, in this study, we employ the simulated annealing ‐long short ‐term memory (SA -LSTM) algorithm and predict optical power fluctuations. First, the raw optical power data are normalized. Subsequently, the SA algorithm is used to perform global optimization of the LSTM algorithm’s key parameters, including the number of iterations, hidden layer units, and learning rate, across a wide range. These optimized parameters are used to construct an LSTM model that can predict future trends in optical power. Experimental results demonstrate that the optical power, predicted using the proposed method with optimal parameters, exhibits a root ‐mean ‐square error of 2.83 × 10.3 dB, which is only 0.14% of the mean value. The prediction accuracy of the proposed method is 91.7%, 91.0%, 96.4%, and 96.3% higher than those of the autoregressive moving -average radial ‐basis ‐function, autoregressive integrated moving ‐average support ‐vector ‐machine,autoregressive moving-average LSTM, and autoregressive integrated moving-average LSTM algorithms, respectively. Thus, the proposed method can predict optical power trends in power cable systems with high accuracy, showing significant potential for engineering applications.

    参考文献
    相似文献
    引证文献
引用本文

陈锦山,吴诗雨,林国栋,王新澜,林树,李兆祥,赵丽娟.基于 SA-LSTM的电力光缆光功率预测方法[J].半导体光电,2025,46(3):536-542. CHEN Jinshan, WU Shiyu, LIN Guodong, WANG Xinlan, LIN Shu, LI Zhaoxiang, ZHAO Lijuan. Simulated Annealing-Long Short-Term Memory Based Light-Power Prediction Method for Power Cables[J].,2025,46(3):536-542.

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-01-22
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-07-28
  • 出版日期:
文章二维码