基于MA-DBN算法的GIS故障预测

GIS fault prediction based on MA-DBN algorithm

对气体绝缘金属封闭开关设备(gas insulatedmetal-enclosed switchgear,GIS)提前进行准确的故障预测,可减少因其故障导致电力系统无法正常运转带来的经济损失。借鉴传统的GIS故障诊断方法,提出采用移动平均模型(moving average,MA)对SF6分解气体成分进行预测,同时结合GIS的气室温度、密度、压力等历史数据,通过深度信念网络(deep belief network,DBN)对预测数据进行误差修正,预测得到GIS未来的气体浓度及故障类型。结果表明:对多台GIS的历史数据及其真实故障信息进行测试,其整体准确率达到了92%,提出的GIS故障预测方法是可行性和有效的。

Accurate fault prediction of gas insulatedmetal-enclosed switchgear(GIS)in advance can reduce the economic loss caused by the failure of power system.Based on the traditional GIS fault diagnosis methods,the moving average(MA)model is used to predict the gas composition of SF6 decomposition.Combined with the historical data of gas chamber temperature,density and pressure of GIS,the prediction data were corrected by deep belief network(DBN)to predict the gas concentration and type of GIS in the future.The results show that after the test of historical data and real fault information of several GIS,the overall accuracy rate reached 92%,which proved the feasibility and effectiveness of the proposed GIS fault prediction method.