基于时域介电法和动态贝叶斯网络的变压器油纸绝缘老化状态评估

为了能够实现对变压器油纸绝缘老化状态的准确评估,文中在极化/去极化电流方法的基础上,提出6种表征油纸老化的特征量,并将动态贝叶斯网络引入到油纸绝缘老化状态评估问题中,建立了相应的老化阶段预测模型,针对该模型进行了实例验证。结果表明:从PDC及拓展Debye模型提取的绝缘纸电量Q(paper)、稳定平均电导电流i^-σ以及平均极化时间常数T^-paper对油纸老化状态反映敏感,且与聚合度有良好的拟合关系,可作为表征绝缘纸老化的特征量;油纸绝缘老化状态评估问题符合动态贝叶斯网络的应用条件,通过划分特征量状态及老化阶段,可以反映变量之间的时序关系;油纸老化状态评估模型验证的准确度为91.3%,平均相对误差为0.108,故验证了动态贝叶斯网络应用于油纸绝缘老化状态评估的准确性和稳定性。

Assessment of Aging State of Transformer Oil-paper Insulation Based on Time Domain Dielectric Method and Dynamic Bayesian Network

In order to make an accurate assessment of the oil-paper insulation aging state of transformers, six characteristic quantities are extracted based on polarization/depolarization current method. Dynamic Bayesian network is introduced into the assessment of oil-paper insulation aging state, and an aging step prediction model that verified by the experiment is established. The results indicate that: the quantity of insulation paper electricity Qpaper,the stable average conduction current i^-σ, and the average polarization characteristic time T^-paper are sensitive to oil-paper insulation aging state, and fit well w.h the degree of polymerization, so that, these can be used as characteristics to represent the aging of insulation paper. The assessment of oil-paper insulation is in accordance with the application conditions of dynamic Bayesian network, by dividing the characteristics state and aging stage, it can reflect the timing relationship between variables. The accuracy of the assessment model of oil-paper insulation aging state is 91.3%, relative average error is 0.108. So the accuracy and stability of applying dynamic Bayesian network to the assessment of oil-paper insulation aging state.