基于改进K-最近邻算法的变电站设备分类识别方法研究

罗金满; 梁浩波; 王莉娜; 刘卓贤; 肖啸

罗金满,梁浩波,王莉娜,刘卓贤,肖啸.基于改进K-最近邻算法的变电站设备分类识别方法研究[J].电测与仪表,2024,61(10):50-56.
Jinman Luo,Haobo Liang,Lina Wang,Zhuoxian Liu,Xiao Xiao.Research on classification and recognition method of substation equipment based on improved K-nearest neighbor algorithm[J].Electrical Measurement & Instrumentation,2024,61(10):50-56.

基于改进K-最近邻算法的变电站设备分类识别方法研究

Research on classification and recognition method of substation equipment based on improved K-nearest neighbor algorithm

DOI：10.19753/j.issn1001-1390.2024.10.007

中文关键词: 三维点云数据变电站设备分类识别 K-最近邻粒子群算法

英文关键词:3D point cloud data, substation equipment, classification recognition, K-nearest neighbor, particle swarm optimization algorithm

基金项目:南方电网公司信息化重点项目（031900HK42200008）

作者中文名	作者英文名	单位
罗金满	Jinman Luo	广东电网有限责任公司东莞供电局信息中心
梁浩波	Haobo Liang	广东电网有限责任公司东莞供电局信息中心
王莉娜	Lina Wang	广东电网有限责任公司东莞供电局信息中心
刘卓贤	Zhuoxian Liu	广东电网有限责任公司东莞供电局信息中心
肖啸	Xiao Xiao	南方电网深圳数字电网研究院有限公司

摘要点击次数: 430

中文摘要:

针对变电站设备三维点云数据采集缺陷造成的场景重建精度低、效率差等问题，在对识别过程进行分析的基础上，提出了一种结合K-最近邻分类算法和改进粒子群算的变电站设备分类识别方法。使用改进的粒子群优化算法来优化K-最近邻分类器的输入权重，提高了设备的分类识别精度。通过仿真进行对比分析，验证该方法的优越性。结果表明，采用该方法的分类识别效果显著，训练准确率达到100％，测试准确率达到99％，与传统识别方法相比，识别准确率从97%提高到99%，平均识别时间从85.81 s降低到0.19 s。该方法解决了变电站设备三维点云数据采集缺陷造成的场景重建精度低、效率差、识别率低等问题，有效提高了变电站设备的分类识别效果，具有良好的实用价值和可操作性。

英文摘要:

Aiming at the problems of low accuracy and poor efficiency of scene reconstruction caused by the defects of three-dimensional point cloud data acquisition of substation equipment, based on the analysis of the identification process, this paper proposes a classification and identification method of substation equipment combining K-nearest neighbor classification algorithm and improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm is used to optimize the input weight of the K-nearest neighbor classifier and improve the classification and recognition accuracy of the equipment. The superiority of this method is verified by simulation and comparison analysis. The results show that the classification recognition effect of the proposed method is remarkable, the training accuracy rate is 100%, and the test accuracy rate is 99%. Compared with the traditional recognition method, the recognition accuracy rate is improved from 97% to 99%, and the average recognition time is reduced from 85.81s to 0.19s. This method solves the problems of low scene reconstruction accuracy, poor efficiency and low recognition rate caused by the defect of three-dimensional point cloud data acquisition of substation equipment, effectively improves the classification and recognition effect of substation equipment, which has good practical value and operability.

查看全文查看/发表评论下载PDF阅读器

关闭