Research On Fault Diagnosis Of Hydroelectric Unit Vibration Based On Variable Precision Rough Set And Neural Network

Along with increasing of unit capacity and enhancing the positional of hydroelectricity in the whole electricity system in our country, the stabilization of hydroelectric units become more and more important to safe operation of whole system. Duo to vibration of hydroelectric units is a main parameter that reflects the running state of the units, it is important to ensure running safe and stable of hydroelectric units for finding out and diagnosing and eliminating system's fault in time.In the paper, according to the present situation and developing trends on condition monitoring and fault diagnosis of hydroelectric units, this dissertation researches into fault diagnosis methods based on Variable Precision Rough Set(VPRS) theory and RBF neural network respectively, then the method based on variable precision rough set theory & RBF neural network is studied, and attempts it to hydroelectric sets vibration fault diagnoses. The main study in the paper is as follows:The mechanism of the major vibration fault and symptom of hydroelectric units, and the methods that can identify unit's vibration fault are analyzed and induced.Based on the sum-up of the fundament principle of Rough set theory, the method of data reduced and rules generated are discussed. The attribute reduction method of variable precision rough set based on genetic algorithm is also studied. The excellent and the defect of method are analyzed by an example of fault diagnosis of hydroelectric units vibration.The fault diagnosis method based on neural network is introduced, based on the analysis of the structure and arithmetic of RBF neural network. The arithmetic and program for fault diagnosis is researched, then the excellent and the defect of method are analyzed by an example of fault diagnosis of hydroelectric units vibration.By analyzing the excellent and the defect of two methods, according as the complexity of vibration failure and the excessive date of monitoring for hydroelectric units, the fault diagnosis method which based on variable precision rough set & RBF neural network is studied. Firstly, the information of hydroelectric units is reduced and the fault rule is generated on variableprecision rough set, then the information is diagnosed by RBF neural network. The method, not only develops fully the reduction ability of rough set theory and the high fault-tolerance ability of neural network, but also gets over the limitations of neural network which is the poor ability of determining the redundancy or usefulness of the knowledge, and decreases the number of the network input nerve cells effectively, and predigests the network structural and improves the accuracy of fault diagnosis, can make up the defect of rough set which is the sensitivity for noise of input information, accordingly improves the accuracy of fault diagnosis. Finally, the result of example proves that the proposed method has high validity and obvious advantage.