基于优化决策树算法的变电站故障诊断系统研究  

Research on Substation Fault Diagnosis System Based on Optimized Decision Tree Algorithm

作　　者：李杰[1] 孙鹤林[1] 雷一鸣 田晓雷 蔡正梓 LI Jie;SUN He-lin;LEI Yi-ming;TIAN Xiao-lei;CAI Zheng-zi(State Grid Beijing Electric Power Company,Beijing 100000 China)

机构地区：[1]国网北京市电力公司,北京100000

出　　处：《自动化技术与应用》2023年第6期112-115,154,共5页Techniques of Automation and Applications

基　　金：国家电网公司科学技术项目(5500-202011091A-0-0-00)。

摘　　要：为有效提升变电站运行期间故障自动化诊断效率和准确率,采用阈值近邻迭代法对样本数量进行优化,同时引入平衡系数对测试属性选择进行优化,然后再采用优化二分离散算法对连续属性离散性进行优化,构建基于优化决策树算法的变电站故障诊断系统。通过优化决策树、ID3以及C4.5三种算法的应用效果对比分析,得出基于优化决策树算法分类准确性更高、决策树构建速度更快、决策树规模更小这一结论,能够显著提升监控系统在变电站运行期间的应用效果,对于变电站长期运行稳定与安全具有重要意义。In order to effectively improve the efficiency and accuracy of automatic fault diagnosis during substation operation,the threshold nearest neighbor iteration method is used to optimize the number of samples,and the balance coefficient is introduced to optimize the selection of test attributes.Then the optimal two separation algorithm is used to optimize the discreteness of continuous attributes,and the substation fault diagnosis system based on optimal decision tree algorithm is constructed.Through the comparative analysis of the application effect of optimized decision tree algorithm,ID3 algorithm and C4.5 algorithm,it is concluded that the classification accuracy of optimized decision tree algorithm is higher,the construction speed of decision tree is faster,and the scale of decision tree is smaller,which can significantly improve the application effect of monitoring system in the operation period of substation,and has important significance for the long-term stability and safety of substation operation.

关 键 词：优化决策树算法 故障诊断系统 阈值近邻迭代法 平衡系数 优化二分离散算法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]