基于图像编码与深度学习的非侵入式负荷识别方法

非侵入式负荷识别(non-intrusive load monitoring,NILM)是一种不依赖用户内部装置,仅凭借外部分析工具和手段即可实现用户用电行为自动感知的方法.提高非侵入式负荷识别的精度,对于开展用能监测服务、实现节能降损具有重要意义.提出了一种基于彩色图像编码与深度学习的电力负荷识别方法.该方法首先在传统电压-电流(V-I)灰色轨迹法的基础上,利用双线性插值技术有效解决了像素点不连续的问题;然后考虑了特征之间的互补性,通过构造电流(R)、电压(G)和相位(B)3个通道,将数值特征嵌入灰色V-I轨迹中,从而得到了蕴含丰富电气特征的彩色V-I图像;最后,采用AlexNet深度学习算法对彩色V-I图像和对应设备标签进行有监督的学习,从而实现了不同类别电器设备的有效辨识.算例测试结果表明,提出的负荷识别方法的准确率高达97.7％.该结果充分验证了上述方法的有效性.

Non-intrusive Load Monitoring Method Based on Color Encoding And Deep Learning

Non-intrusive load monitoring(NILM) is a method that does not rely on the user''s internal devices, but can realize the automatic perception of the user''s electricity consumption behavior with only external analysis tools and methods. Improving the accuracy of NILM is of great significance to the development of energy consumption monitoring services and the realization of energy conservation. In this paper, a power load identification method based on color image coding and deep learning algorithm is proposed. Firstly, based on the traditional gray V-I trajectory method, bilinear interpolation technology is used to deal with the problem of pixel discontinuity. Then, considering the complementarity between features, the numerical features are embedded into gray V-I trajectory by constructing three channels of current-R, voltage-G and phase-B, and the color V-I image with rich electrical information is obtained successfully. Finally, the deep learning algorithm AlexNet is used to learn color V-I images and equipment labels in a supervised way, so as to realize the effective identification of different types of electrical equipment. The test results show that the accuracy of the proposed load identification method is as high as 97.7%. The results fully verify the effectiveness of the above method.