| 引用本文: | 程志友,胡乐乐,陈思源,等.基于多V-I轨迹融合的非侵入式负荷识别方法[J].电力系统保护与控制,2025,53(11):63-71.[点击复制] |
| CHENG Zhiyou,HU Lele,CHEN Siyuan,et al.Non-intrusive load identification method based on multiple V-I trajectory fusion[J].Power System Protection and Control,2025,53(11):63-71[点击复制] |
|
| 摘要: |
| 在负荷识别领域中，仅使用单一负荷特征难以有效区分相似轨迹的负荷。为解决这一问题，提出了一种基于多V-I(电压-电流)轨迹融合的非侵入式负荷识别方法。该方法首先对高频采样数据进行预处理，从中提取基波电压(V1)、基波电流(I1)以及最大谐波电流(Ih max)。随后使用基波电压分别与基波电流和最大谐波电流相结合，构建了V1-I1轨迹和V1-Ih max轨迹。最后将这两种轨迹特征输入到二维卷积神经网络(2D convolutional neural network, 2D-CNN)中进行负荷分类，通过PLAID和WHITED两个公共数据集进行验证，所提出的负荷识别方法的准确率高达99.66%和99.81%。该实验结果表明，所提方法不仅增加了信息量，还提高了负荷识别的准确率，在实际电力监控和负荷管理中具有应用价值。 |
| 关键词: 非侵入式负荷识别 相似轨迹 V1-I1轨迹 V1-Ih max轨迹 卷积神经网络 |
| DOI：10.19783/j.cnki.pspc.240660 |
| 投稿时间：2024-05-28修订日期：2024-07-31 |
| 基金项目:国家自然科学基金项目资助(6227020935)；安徽省科技重大专项资助(18030901018) |
|
| Non-intrusive load identification method based on multiple V-I trajectory fusion |
| CHENG Zhiyou1,HU Lele2,CHEN Siyuan2,YANG Meng2 |
| (1. Internet School, Anhui University, Hefei 230039, China; 2. School of Electronic and Information Engineering,
Anhui University, Hefei 230601, China) |
| Abstract: |
| In the field of load identification, it is difficult to effectively distinguish loads with similar trajectories using a single load feature. To address this issue, a non-intrusive load identification method based on multiple V-I (voltage-current) trajectory fusion is proposed. This method first preprocesses high-frequency sampling data to extract the fundamental voltage (V1), fundamental current (I1), and maximum harmonic current (Ih max). Subsequently, the fundamental voltage is combined with the fundamental current and maximum harmonic current to construct V1-I1 trajectories and V1-Ih max trajectories. Finally, these two trajectory features are input into a two-dimensional convolutional neural network (2DCNN) for load classification. Validation using the public PLAID and WHITED datasets shows that the proposed load identification method achieves accuracies of 99.66% and 99.81%, respectively. These results indicate that the proposed method not only enriches the information used for classification but also significantly improves load identification accuracy, demonstrating its practical application value in power monitoring and load management. |
| Key words: non-intrusive load identification similar trajectories V1-I1 trajectory V1-Ih max trajectory convolutional neural network |
