The Study Of The Generator Rotor Winding Short Fault Diagnosis Based On Fuzzy Neural Networks

The turbo-generator is a kind of synchronous electrical machine rotating at high speed. During its operation, the rotor of the machine is always under the great suppression of centrifugal force. The surface of rotor is the most important part to enduring the forces and thermal energy. So the windings of rotor are in a common problem of happening one-point to earth short fault, two-point to earth short fault, or turn-to-turn short fault. Because the fault of field windings turn-to-turn shorted do not cause much effect of the operation of the machine or its fault characteristics are not very clear, most of the faults are neglected. But if a machine is in this kind state in a long periods of time, it may cause other faults such as one-point to earth short fault or two-point to earth short fault. Because the resistant of one-point to earth short fault is so great that it may not cause much trouble except affecting the rotor's using periods of time. The two-point to earth short fault may cause short currents which can lead to serious accident. So it is necessary to monitor the state of field windings of turbo-generator and detect the kind of fault and predict its inclination. Thus the security level of power systemwill be improved. In this paper the technologies of model forecasting and the fault model identification and the intelligence monitoring using fuzzy system and neural network are studied.First, the 2-D model for analyzing is provided. According to the theory of Ward, the field of electrical magnetism is formed under the base of linearization. Then the field model in reality will be formed if the one without fault is added. According to the model, the harmonic voltages at the frequency of 1 OOHz at theside of stator are analyzed. Because there will be no such harmonic voltages in the normal state, it can be treated as a fault characteristics. At the same time the relationship between the current of field and generator output values and its parameter is discussed. It is the current that produce the magnetism field, so it will be an important factor in detecting the faults of rotor windings.Second, the forecasting model for the field current based on the fuzzy system is discussed. The fuzzy system is founded to describe the operation mechanisms of human mind. It can'describe a function founded on a set especially with non-linear characteristics with high accuracy. Because it has the non-linear characteristic. It is another model simulation application just as neural networks and polynomial theory. In this paper the predictive system of field current is studied using Takagi-Sugeno fuzzy system. The input parameters are drawn from the turbo-generator's stator. The fuzzy system is constructed through the artificial neural learning algorithm.Third, the fault diagnosis using neural networks is discussed in this paper, especially the internal backward neural network with deviation elements. Its model and learning algorithm are showed in detail. Then the fuzzy neural network is formed. This kind of system integrates the characteristics of the fuzzy logic system and neural network. It can complete the logic deduction procession using the theory of neural computing.Last, the different functional parts of the diagnosis system are discussed. The simulation results show it has a satisfied capability in diagnosis the faults in the rotor windings of the two poles turbo-generator.