基于改进模糊时间序列的变压器油中气体预测方法

Prediction of dissolved gas concentration in oil based on improved fuzzy time series

对变压器油中溶解气体含量进行预测有助于及早发现变压器内部的潜伏性故障,且对于更好地实现状态检修有着重要的指导意义.针对变压器油中气体组分数据丰富、正常运行状态下各组分含量变化趋势不明显的特点,提出基于模糊时间序列模型的变压器油中气体组分预测方法.考虑到油中气体组分的变化是相互作用和影响的,从论域划分角度对经典模糊时间序列模型进行改进,提出基于空间模糊C均值(fuzzy C-means,FCM)论域划分的多因素模糊时间序列模型.通过实例分析证明该方法能很好地拟合油中气体组分的变化趋势,且与经典模糊时间序列模型及一维FCM划分模糊时间序列模型的对比分析,验证了改进模型在预测效果上的优越性.

The prediction of dissolved gas content in transformer oil is helpful for early detection of latent faults in transformer, and it has important guiding significance for better condition based maintenance. In view of the abundant data of transformer dissolved gas analysis （DGA）, and that the trend of the change of dissolved gas content in oil under normal running condition is not obvious, a prediction method based on fuzzy time series model is proposed. Considering that the change in dissolved gas content in oil is interac- tion and influenced, in this paper, the classical fuzzy time series model is improved from the view of domain division; and a multifactor fuzzy time series model is proposed based on spatial fuzzy C-means （FCM） domain partition. The case study shows that the method can well fit the changing trend of DGA data, and compared with the classic fuzzy time series model and the one-dimensional FCM fuzzy time series model, the superiority of the improved model in prediction is verified.