The Application Of Condenser Fault Diagnosis Based On Rough Set And Ensemble SVM

As an important part of power plant operation,condenser mainly undertakes to condense steam turbine exhaust into water and maintain the vacuum needed.In the actual operation of the power plant,due to the influence of various factors,the operation of the equipment of each power plant may appear unstable,which may cause the malfunction of the equipment and even cause the unit to run abnormally.Therefore,it is very important for the safe and stable operation of the unit to monitor the operation State of condenser accurately and to find out and eliminate the unstable factors which affect the condenser fault.In this paper,the main factors affecting the condenser vacuum drop are analyzed and the symptom database of condenser failure is established based on the analysis of the operation characteristics of condenser.Then,the rough set theory and support vector mechanism theory are studied in detail and analyzed with concrete examples from the point of view of combining theory with practical application.Based on the method of combining rough set and integrated LS algorithm,the vacuum fault diagnosis model of condenser is established.The specific work is as follows:In view of the high input dimension of the sample set of condenser fault symptom,the problem of complex structure and poor real-time variation of the classifier is caused.According to the attribute reduction feature of rough set,the sample set of rough set is used to keep the classification ability of the original symptom sample set unchanged.In view of the error rate of a single support vector machine due to insufficient training or the parameter setting problem,the Ensemble learning algorithm is introduced,and the SVM Ensemble entity is produced by using different kernel functions and parameters.After selecting the fault symptom space subset which is produced by rough and intensive simplification,the integrated individual with large difference is constructed and the SVM is trained,and finally the final diagnosis result is generated by voting mechanism,which further increases the generalization ability of the whole model and improves the performance of the system.