基于RPROP神经网络算法的主变DGA故障诊断模型

故障诊断模型是开展输变电设备状态检修的核心环节之一,文中采用弹性反馈(RPROP)神经网络算法建立主变压器油中溶解气体的神经网络故障诊断模型,通过与带动量因子的标准反向传播(BP)算法、Bold Driver算法、SuperSAB算法相比较,表明了RPROP算法在故障模式识别中具有更好的学习效率与泛化能力,故障诊断的准确度高于传统分析方法,在变电设备状态诊断中具有良好的应用前景.

APPLICATION OF RPROP ANN BASED FAULT DIAGNOSIS MODEL FOR TRANSFORMER DISSOLVED GAS-IN-OIL ANALYSIS

Fault diagnosis model is one of the core algorithms in transmission and transformation equipment. An artificial neural network (ANN) programming based on RPROP (resilient propagation ) algorithm is presented for the dissolved gas-in-oil analysis (DGA) methodologies. The comparative analysis shows RPROP algorithm provides both higher learning efficiency and stronger generalization capacity versus standard BP, Bold Driver and SuperSAB algorithms once used in DGA. When RPROP is applied to the transformer DGA, the fault diagnosis accuracy is evidently enhanced compared to other conventional methods. Therefore, it shows a promising future in the diagnostic field for power transformation equipment.