基于深度卷积神经网络的电力系统故障预测
DOI:
作者:
作者单位:

1.国网山西省电力公司电力调度控制中心;2.哈尔滨工业大学电气工程系;3.南瑞集团国网电力科学研究院有限公司

作者简介:

通讯作者:

基金项目:

国家重点研发计划项目(多能源电力系统互补协调调度与控制,2017YFB0902200);国网山西省电力公司科技项目WD160274(特高压直流及多电压等级交流外送通道的新能源并网智能控制关键技术研究与示范)


Power system fault prediction based on deep convolutional neural networkZHU Yanfang 1 YAN Lei 1* CHANG Kang2,3 ZHAO Wenna1 LI Yuan1 XU Limei1
Author:
Affiliation:

DispatchingandControlCenterofStateGridShanxiElectricPowerCompany

Fund Project:

National key R & D projects (complementary and coordinated dispatching and control of multi energy power system, 2017yff0902200); State Grid Shanxi Electric Power Co., Ltd. science and technology projects WD160274 (Research and demonstration on Key Technologies of new energy grid connected intelligent control of UHVDC and multi voltage level AC transmission channel)

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

    通过对深度卷积神经网络的深入研究,提出基于深度卷积神经网络的电力系统故障预测方法,保障系统安全运行。采用广域测量系统测量每个支路与节点,将获得的功率与关键特征值分别作为深度卷积神经网络模型输入、输出,训练这两个数据,并使用深度卷积神经网络AlexNet分析输入数据与输出数据的映射关系,建立基于深度卷积神经网络的电力系统故障预测模型,通过特征值分组、振荡模式筛选、数据预处理、模型训练和模型评估,实现电力系统运行状态评估,完成电力系统故障预测。实验结果说明:该方法的关键特征值计算结果与实际结果基本一致,可靠性高;使用正则化可提升模型泛化效果,防止模型过拟合;准确率指标和综合评价指标最高,评估效果优势显著。

    Abstract:

    After in-depth research on deep convolutional neural networks, a power system fault prediction method based on deep convolutional neural networks is proposed to ensure the safe operation of the system. Use a wide-area measurement system to measure each branch and node, and use the obtained power and key feature values as the input and output of the deep convolutional neural network model, train these two data, and use the deep convolutional neural network AlexNet to analyze the input data The mapping relationship with the output data, the establishment of a power system fault prediction model based on deep convolutional neural networks, through feature value grouping, oscillation mode screening, data preprocessing, model training and model evaluation, the power system operation status evaluation is realized, and the power system is completed Failure prediction. The experimental results show that the calculation results of the key feature values of this method are basically consistent with the actual results, and the reliability is high; the use of regularization can improve the generalization effect of the model and prevent the model from overfitting; the accuracy index and comprehensive evaluation index are the highest, and the evaluation effect is superior Significantly.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
历史
  • 收稿日期:2021-10-14
  • 最后修改日期:2022-03-18
  • 录用日期:2022-01-18
  • 在线发布日期: