Fuzzy neural classifier for fault diagnosis of transformer based on rough sets theory

Due to enduring more disturbance such as environment varieties and surveying interference and information transmission mistakes as well as arisen error while processing data in surveying and monitoring state information of transformer, thus uncertain and incomplete information and ill data may be produced. So the study how to apply these data to achieve the approving effect is a very significant job for fault diagnosis of transformer. Moreover, real time is another important characteristic so as to meet high-speed diagnosis requirements. Based on points, a fuzzy neural classifier is proposed based on rough sets theory in this paper, the method firstly considers all sorts of gas capacities in transformer oil to form Rogers ratio diagnosis table, then rough sets is applied to implement attributes reduction and a simplified decision table is got, fuzzy algorithm with Gauss subjection function makes attribute values fuzzy, afterwards, fuzzy attributes are connected to input neurons of neural classifier to make patterns classified, finally, a fuzzy neural classifier is formed for fault diagnosis for transformer. The practical results show the approach can effectively minimize the problem-solving scale and improve real time properties, and owns high anti-inference capabilities, and is an effective method for fault diagnosis of transformer