基于K-means聚类算法的智能照明系统设计与开发  

Design and Development of Intelligent Lighting System Based on K-means Clustering Algorithm

作　　者：高杰 魏强 GAO Jie;WEI Qiang(College of Science,Chongqing University of Technology,Chongqing 401135,China)

机构地区：[1]重庆理工大学理学院,重庆401135

出　　处：《中国照明电器》2022年第3期28-35,共8页China Light & Lighting

基　　金：重庆市教委科学技术研究计划项目(KJQN201901134)。

摘　　要：本文提出了一种基于K-means聚类算法的智能照明系统。系统采用MQTT(消息队列遥测传输)通信协议传输控制信号和采集用户行为习惯数据,选取用户的日常用灯时间和用灯时长作为两个特征值。为了使算法适应不同的用户数据,提出了一种确定K-means聚类算法的聚类数搜索范围上下界的方法,并通过选择有效性指标,以及基于用户行为习惯数据设定初始中心点,确定最佳聚类数。实验利用ESP8266模块、MQTT服务器和手机端测试界面测试灯的控制和数据采集,借助PyCharm软件对传统K-means算法和改进后的K-means算法在用户用灯行为习惯数据上进行聚类数搜索范围的确定、有效性指标和聚类效果比较。实验结果表明,采用MQTT通信协议能够控制灯的状态,平均响应时间为105 ms,改进的K-means聚类算法能够很好地对用户用灯行为习惯数据进行聚类,自动为用户提供正确的开灯时间和开灯持续时间。This paper presents an intelligent lighting system based on K-means clustering algorithm. The system uses MQTT(Message Queuing Telemetry Transport) communication protocol to transmit control signals and collect user behavior habit data, and selects the user’s daily light usage time and time of duration as two characteristic values. In order to adapt the algorithm to different user data, a method to determine the upper and lower bounds of the search range of the clustering number of K-means clustering algorithm is proposed, and the optimal clustering number is determined by selecting the effectiveness index and setting the initial center point based on the user behavior habit data.The experiment uses the ESP8266 module, MQTT server and mobile terminal test interface to control and collect data from the lamp.With the help of PyCharm software, the traditional K-means algorithm and the improved K-means algorithm are compared in the determination of the search range of clustering number, effectiveness index and clustering effect on the user’s lamp behavior habit data.The experimental results show that MQTT communication protocol can be used to control the status of the lamp, and the average response time is 105 ms, and the improved K-means clustering algorithm can cluster the data of users’ lighting behavior habits, and automatically provide users with the correct lighting time and duration.

关 键 词：机器学习 K-MEANS 初始中心点 用户行为习惯 最佳聚类数 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]