基于一维卷积神经网络的家庭用户特征识别方法

打开文本图片集

中图分类号：TB9;TM933 文献标志码：A 文章编号：1674-5124（2025）06-0025-06

Household characteristics identification method based on one-dimensional convolution neural network

XU Jihe¹，ZHU Liang²，YAN Yi²，ZHOUJianan³，WENHe³ （1. Pingxiang Power Suply Company， State Grid Jiangxi Power Company，Pingxiang 330ooo， China; 2.Power Supply Service Management Center，State Grid Jiangxi Power Company，Nanchang 33o077，China; 3.College ofElectrical and Information Engineering， Hunan University， Changsha 41oooo， China）

Abstract： Users’electricity consumption datasets provided by smart energy meters can reflect the users ’ electricity consumption characteristics，which provides a basis for analyzing the household characteristics. Aimingat the effcient classification of household characteristics，this paper studies a household characteristics identification method based on one-dimensional smart energy meter electricity consumption data. In this paper， a one-dimensional convolution neural network suitable for the time series data of smart electric energy meter is designed. Taking the user's electric energy consumption data （one-dimensional time series） measured by the smart electric energy meter as the input， and the pooling layer is removed after the first two convolution operations of the network to achieve the preservation of early features and to achieve accurate classification of the household characteristics.In order to prove the effectiveness of the method proposed in this paper，this paper conducts comparative experiments on public datasets.The experiments show that our method achieves 55%～78% accuracy in the classification of several the household characteristics.

Keywords： deep leaming; one-dimensional convolution neural network; classification; household characteristics; smart energy meters

0 引言

家庭用户特征，包括用户的年龄、薪资、房屋状况、社会关系等，可以帮助零售商了解不同用户的生活习惯和用电模式，有助于公用事业和零售商实施更有效的需求响应方案和更个性化的服务，并就需求响应和能源效率计划的目标做出更可靠的决策。（剩余9656字）