Fault Diagnosis of Rotor Inter-turn Short-circuit Based on Self-adaptive Radial Basis Function Network

As the structure of generator is complex, and the saturated magnetic field is nonlinear, it is difficult to use generator´s mathematic model to calculate field current directly, and then hard to use the deduced method to identify rotor interturn short-circuit fault, the computation time is long, and the computation precision is not good, so it is difficult to diagnose online. We proposed a new radial basis function network (RBFN) in this paper, it is two levels iteration clustering algorithm. This method overcomes the shortcomings that other algorithm have, it can compute the nerve unit number, center and width of RBSN implication layer automatically, and the clustering result is more reasonable, consequently increase the precision of RBSN. In condition of not adding any equipment, using actual measured inactive power, active power and terminal voltage etc. electric parameters of QFSN-300-2 generator as input, and in virtue of self-adaptive RBSN to identify generator field current, combining rotor inter-turn short-circuit fault diagnosis model to identify rotor inter-turn short-circuit fault online, it is faster than simply use generator mathematic model, and has high precision. After certification, the result is: comparing to the traditional diagnosis model, it increases the diagnosis precision and speed greatly, and satisfies the need of online identification