摘要
为了对电力变压器在出现异常征兆时基于色谱数据进行短期预测,且在可能存在多重故障类型时能有效诊断,在变压器油色谱分析时引入云理论并进行相应改进。利用云变换算法将故障变压器油色谱数据转换成符合人认知的多个定性云概念,并提出发掘油中气体云概念与故障类型间关系的云推理机制。基于分析油中气体单个检修周期内的变化规律,利用云理论对短期内油中气体变化的期望值进行预测,然后利用改进的云推理预测组合规则发生器推理得到一系列有稳定倾向的故障预测结果,并求解相应的可信度,最终给出可信度大于设定阈值的若干预测结果供选取。多实例分析验证表明,云推理故障诊断能对变压器各故障类型及多重故障准确诊断;云预测模型能在非等间隔时间的数据序列下,对适当波动的油中溶解气体分析(DGA)数据准确预测其短期变化趋势及期望值。
With the introduction of cloud theory, we made improvements in analyzing transformer oil dissolved gas to forecast short-term faults for power transformer when abnormal signs appeared and to realize fault diagnosis with cloud reasoning when multiple faults existed. Sample data of dissolved gas in transformer were transformed into multiple qualitative cloud concepts through a cloud transform algorithm. Then the transformer fault diagnosis could be performed with the cloud inference mechanism connecting the cloud concepts to fault type. On the basis of studying the variation of dissolved gas in a single maintenance cycle and cloud theory, we could predict the expected value of dissolved gas oil in a short term. Furthermore, using an improved combined rule generator of cloud-inference prediction, we deduced a series of failure predictions that tended to be stable, calculated their reliability, and eventually obtained several predictions with enough reliability as diagnosis results. Further analysis of multiple instances indicate that the proposed cloud inference analysis can diagnose both single-type and multiple-type transformer faults accurately. Moreover, with unequal interval data sequence, the cloud model can accurately predict short-term trends and expectations of data by reasonably fluctuating dissolved gas analysis（DGA）.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2014年第5期1453-1460,共8页
High Voltage Engineering
基金
国家重点基础研究发展计划(973计划)(2009CB724505)
国家创新研究群体基金(51021005)~~
关键词
云理论
电力变压器
故障诊断
短期预测
云变换
关联规则
cloud theory
power transformer
fault diagnosis
short-term prediction
cloud transform
association rules
