摘要
非侵入式负荷监测可以在减少检测装置数目的前提下,获取用户家用电器的启停和用电信息,然而由于技术性问题导致辨识过程较为复杂、特征不易被提取。本文提出将暂态的有功、无功功率的轨迹曲线进行绘制,利用拉格朗日插值法插补两个像素之间的间断点,防止因采样时间而造成辨识结果的不同。通过改进的Lenet-5卷积神经网络来辨识功率曲线的特征,能够在线辨识出负荷的种类。通过辨识仿真用电数据和实验数据对比可知,该算法辨识出的负荷具有一定的精度,为该类应用提供了参考。
Non-intrusive load monitoring can reduce the number of detection devices and derive the start-stop and power consumption information of the user’s household appliances.However,due to technical problems,the identification process is more complicated and features are difficult to extract.This paper proposes to draw transient active and reactive power trajectory curves,and the Lagrangian interpolation is used to interpolate the discontinuity between two pixels to prevent the difference in identification results due to sampling time.Through the improved Lenet-5 convolutional neural network to identify the characteristics of the power curve,the type of load can be identified online.By identifying the simulated electricity consumption data and compared with the experimental data,it can be known that the load identified by the algorithm has a certain accuracy,which provides the reference for this kind of applications.
出处
《日用电器》
2021年第2期69-74,共6页
ELECTRICAL APPLIANCES
基金
河北省重点研发计划项目(20311801D)
河北省高层次人才资助项目(A201905008)。
关键词
非侵入式负荷监测
卷积神经网络
拉格朗日插值法
non-invasive load monitoring
convolutional neural network
Lagrangian interpolation
