CAJ | 학술논문

传统的人工智能方法对变压器大量的不完备故障信息不能有效地分析,或在故障数据的离散化过程中由于区间分割不当而无法正确诊断故障甚至误诊。为此,提出了一种基于改进人工鱼群优化粗糙集的变压器故障诊断方法。该方法首先将变压器溶解气体分析（DGA）的值作为条件属性,将故障类型作为决策属性,建立故障决策表,利用鱼群的聚群寻优行为对决策表中的连续属性数据进行离散化;然后采用粗糙集理论对离散化后的决策表进行约简,建立故障诊断规则决策表,大大简化了决策表属性约简的难度,使诊断变得更加简便。最后通过实例验证表明：该方法能够有效地对样本进行离散和约简,与传统方法相比,提高了故障诊断的正确率。
전통적인공지능방법대변압기대량적불완비고장신식불능유효지분석,혹재고장수거적리산화과정중유우구간분할불당이무법정학진단고장심지오진。위차,제출료일충기우개진인공어군우화조조집적변압기고장진단방법。해방법수선장변압기용해기체분석（DGA）적치작위조건속성,장고장류형작위결책속성,건립고장결책표,이용어군적취군심우행위대결책표중적련속속성수거진행리산화;연후채용조조집이론대리산화후적결책표진행약간,건립고장진단규칙결책표,대대간화료결책표속성약간적난도,사진단변득경가간편。최후통과실례험증표명：해방법능구유효지대양본진행리산화약간,여전통방법상비,제고료고장진단적정학솔。
Facing a large number of incomplete fault data, the traditional artificial intelligence methods cannot effectively and timely analyze or accurately diagnosed because of the ill-conditioned problem caused by inefficient discretization approaches. We presented a method based on rough set theory integrated with improved artificial fish swarm algorithm {AFSA} for fault diagnosis of transformer. Firstly, the values of dissolved gas analysis {DGA） in oil were taken as conditional attributes and the type faults were taken as decision attributes. Various relations between fault and symptom were connected, and decision table was established. Then, the improved artificial fish swarm algorithm was used to discrete continuous attribute, and the rough set theory was used to reduce the decision table. Finally, the simplified decision rules were got, which greatly simplified the difficulty of diagnosis. The experimental results indicate that the method increases the diagnosis accuracy compared with the traditional algorithm.