| 引用本文: | 罗斌,王慧芳,倪旭明,李佶,吴昊,林恺丰.基于1D-YOLO网络的电能质量复合扰动检测分类和时间定位模型[J].电力自动化设备,2025,45(6):182-190. |
| LUO Bin,WANG Huifang,NI Xuming,LI Ji,WU Hao,LIN Kaifeng.Recognition and time interval identification model for multiple power quality disturbances based on 1D-YOLO networks[J].Electric Power Automation Equipment,2025,45(6):182-190. |
|
| 摘要: |
| 为准确识别电能质量扰动类型，进一步挖掘扰动内部特征，提出基于一维YOLO(1D-YOLO)网络的电能质量复合扰动检测分类和时间定位模型。模型利用带残差结构的一维卷积神经网络对扰动波形进行深度的特征提取；利用特征金字塔网络融合不同尺度的特征；在3个不同尺寸的特征图上完成8种基本扰动分层检测。以基本扰动为检测对象，有效实现了单一扰动和多重扰动的层次化建模，仿真实验表明模型具有较高的分类准确率和时间定位精度，现场数据进一步证明了模型的泛化性和有效性。 |
| 关键词: 电能质量扰动 时间定位 多任务学习 卷积神经网络 目标检测 |
| DOI：10.16081/j.epae.202503029 |
| 分类号:TM711 |
| 基金项目:国网浙江省电力有限公司科技项目(5211JH24000A) |
|
| Recognition and time interval identification model for multiple power quality disturbances based on 1D-YOLO networks |
|
LUO Bin1, WANG Huifang1, NI Xuming2, LI Ji2, WU Hao2, LIN Kaifeng2
|
|
1.College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China;2.Jinhua Power Supply Company, State Grid Zhejiang Electric Power Co.,Ltd.,Jinhua 321000, China
|
| Abstract: |
| In order to accurately identify the types of power quality disturbances(PQDs) and further mine the internal features of the disturbances, a recognition and time interval identification model for multiple PQDs based on one-dimensional YOLO networks(1D-YOLO) is proposed. The model uses a one-dimensional convolutional neural network(1D-CNN) with residual structure to extract deep features of the disturbance waveform, then merges features of different scales using a feature pyramid network(FPN),and completes the hierarchical detection of eight basic PQDs on three feature maps of different sizes. The hierarchical mode-ling of single and multiple PQDs is effectively achieved by taking the basic PQDs as the detection object, and simulation experiments show that the model has high classification accuracy and time interval identification precision, and data from deployment further prove its generalization and effectiveness. |
| Key words: power quality disturbances time interval identification multi-task learning convolution neural network object detection |
