基于多预测模型融合的电力变压器安全预判  

Safety Prejudging Method for Power Transformer Based on Multi-Prediction Model Fusion

作　　者：李典阳 张育杰 王善渊 冯健[1] 王洪哲 秦领 LI Dianyang;ZHANG Yujie;WANG Shanyuan;FENG Jian;WANG Hongzhe;QIN Ling(College of Information Science and Engineering,Northeastern University,Shenyang 110819,China;State Grid Shenyang Electric Power Co.,Ltd.,Shenyang 110004,China)

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819 [2]国网辽宁省电力有限公司,辽宁沈阳110006

出　　处：《中国电力》2020年第1期72-80,共9页Electric Power

基　　金：国家自然科学基金资助项目(61673093)~~

摘　　要：为解决传统分析措施未能将多源异类的数据信息纳入电网设备故障征兆分析及基于小样本单一算法难以较好应对多种故障类型设备诊断的问题,构建一种不同类型电网设备相关数据的统一化、标准化方法。采用卡方分布算法通过对数据相关性挖掘进行特定故障类型征兆集的选择,以避免层次分析法、格林兰验证等故障征兆集分析方法存在人工经验干扰。构建了一种多算法融合决策方法来避免单一算法决策的弊端。通过实例验证了对电力设备的单一故障类型寻找故障征兆集比针对设备选择故障征兆集具有更好的简约效果与预判准确率。实例还验证了所提融合算法效果好于单一算法。Fault diagnosis and pre-judgment of power grid equipment is an important guarantee for safe operation of power grids.There are many related factors for grid equipment faults.The conventional analytical measures have not considered integrating multisource heterogeneous data into grid equipment fault cause analysis,and the small sample-based unitary algorithm cannot well deal with diagnosis of multi-type fault equipment.A unified and standardized method is presented in this paper for relevant data of different types of power grid equipment.To avoid the artificial experience interference of such fault cause analysis methods as the analytic hierarchy process and Greenland verification,the Chi-Square distribution algorithm is used to select the specific-type fault cause set through mining data correlation.A new multi-algorithm fusion decision method is proposed to avoid the drawback of unitary algorithm decision.It is verified through case study that the proposed fuse algorithm is better than the unitary algorithm in simpleness and pre-judgment accuracy.

关 键 词：电力数据 数据规范化 征兆集选择 融合决策 事件模型 

分 类 号：TM41[电气工程—电器]