信息缺失下变压器状态预警方法

Early-warning Methods of Transformer Conditions Under Information Deficiency

为了解决变压器状态信息间的弱关联问题,从而准确预警变压器的异常状态,引入信息决策理论构建变压器信息决策表,并提取了变压器状态决策规则。通过信息决策表中各决策规则之间的近似关系,提出了概率化表征变压器状态的方法,实现了信息缺失下的变压器状态预警。此外,基于属性依赖度理论,提出了结合遗传算子的最优属性核计算方法,该方法通过提取分析信息决策表的关键条件属性,完成了变压器状态预警规律的总结和发现。采用64个案例验证了所提方法的有效性,并示例说明了变压器状态预警规律的发现过程。研究结果表明:状态预警方法的正判率约为81.3%,该方法能够在信息缺失下预警变压器状态,并具有自学习优化能力,预警精确度会随状态信息的增多而不断提升。变压器状态预警方法能够为变压器智能化运维和风险管控提供理论指导。

In order to solve the problem of weak correlations between transformer state information for accurately early-warning of transformer abnormal conditions,the information decision-making theory is introduced to establish the transformer decision-making table and the decision rules of transformer states are extracted.Based on the correlation analysis of decision rules in the decision-making table,we design an early-warning method of transformer conditions to prognosticate the possible fault types and to calculate their probability,which addresses unreliable early-warning under deficient information.Furthermore,based on the attributes dependency theory,the calculation method of optimal attribute core is proposed via combining genetic operators.The calculation method is utilized to extract the importantly conditional attributes for achieving the summary&discovery of the early-warning rules of transformer conditions.The effectiveness of the proposed early-warning method is verified through 64 cases and the discovery processes of the early-warning rules of transformer conditions are provided by case studies.The results show that the total recognition accuracy of the early-warning method in this paper is 81.3%.The proposed method is capable of early-warning transformer conditions under information deficiency,and has the ability of self-improving the early-warning accuracy by state information gradually accumulated.The proposed methods can facilitate the intelligent operation&maintenance and risk management of power transformers.