神经网络和ds证据信息融合在变压器故障诊断中的应用.pdf

神经网络和DS证据信息融合在变压器故障 诊断中的应用 陈新岗， 田晓霄 重庆理工大学，重庆 400050 Neural Network and DS Evidence Fusion Algorithm for Power Transformer Fault Diagnosis Xingang Chen, Xiaoxiao Tian Chongqing university of technology, Chongqing 400050, China ABSTRACT: As the transformers fault types and 是国内外学者研究的热点。故障的多样性、 fault positions are complexity, complementarity, 不确定性和各故障之间联系的复杂性构成 redundancy and strong characteristics of 了故障诊断技术上的难点，仅靠单一的故障 uncertainty, a satisfactory diagnosis of transformer 特征和诊断方法无法完成诊断任务。因此很 fault types and fault position may not be obtained 多学者将信息融合技术引入到变压器故障 if only one technique is used. In this paper, a 诊断领域 [1]。 synthetic diagnosis method using neural network 本文应用信息融合原理，结合神经网络 and DS evidence theory for transformer fault 和DS方法各自的优点，提出了一种多传感器 diagnosis is presented, combining DGA data and 信息的两级融合结构分别对变压器故障类 application of data fusion theory. This method has 型和故障部位进行诊断，即利用神经网络实 advantage of both neural and DS evidential theory, 现多传感器的初级信息融合，并将神经网络 in theory it can effectively solve the information 的输出作为证据再利用证据理论进特征决 fusion method the problem of uncertainty, and 策。 improve the fault diagnosis system’s accuracy and reliability. A simulation example in this paper 2 信息融合中的神经网络 illustrates that the method combing of neural 信息融合是指将各传感器的冗余和互 network and DS evidence theory is indeed greatly 补信息进行融合，得到被观测对象更加精确 improved integration of the credibility of the data, 的评估,以便对事物进行正确的判断和决 making diagnostic systems easy to design, high 策。目前信息融合算法大致可以分为两类： precision and prone to operation. It can fulfill