基于NRS和D-S证据理论的变压器油纸绝缘状态评估模型

Evaluation model of transformer oil-paper insulation based on neighborhood rough set and dempster-shafer evidence theory

有效的变压器油纸绝缘状态评估方法对制定合理的检修计划有指导意义。为准确诊断油纸绝缘状态,提出基于邻域粗糙集(NRS)与D-S证据理论的评估模型。模型的构建是基于回复电压法多特征量评估数据库,采用证据融合理论对绝缘状态命题进行置信度计算,克服了单一特征量的片面性。同时,为减少主观因素的影响,根据NRS的上下近似定义,提出了基于数据库的基本信度分配方法。针对特征量在反映绝缘状态上的灵敏性不同,通过组合赋权法综合了基于属性重要度的客观权重和专家经验的主观权重,加强了权重分配的合理性。最后对多台未在数据库的不同糠醛含量变压器实测数据进行算例验证,结果表明,置信度在绝缘状态命题的分布不仅能准确反映变压器实际绝缘状态,并能体现其劣化趋势,为检修策略的制定提供参考依据。

An effective assessment method of transformer oil-paper insulation is of guiding significance for making a reasonable maintenance plan.In order to accurately diagnose the oil-paper insulation state,this paper proposes an evaluation model based on neighborhood rough set(NRS)and D-S evidence theory.The construction of the model is based on the multi-characteristics evaluation database of the recovery voltage method,and the evidence fusion theory is used to calculate the confidence of the insulation state proposition,which overcomes the one-sidedness of the evaluation method using single characteristic.At the same time,according to the definition of upper and lower approximation in NRS,a basic reliability allocation method based on database is proposed to reduce the influence of subjective factors.In view of the different sensitivities of the feature quantity in reflecting the insulation state,the objective weight based on attribute importance and the subjective weight based on expert experience are synthesized through the combination weighting method,which strengthens the rationality of weight distribution.Finally,the measured data of multiple transformers with different furfural content which are not in the database are calculated by examples.The results show that the distribution of confidence in the insulation state proposition accurately reflects the actual insulation state of the transformer,and its deterioration trend,and provides a reference for the formulation of maintenance strategies.