Condenser Fault Diagnosis System Based On Neural Network

Working medium in the operation of the plant state changes follows the Rankine cycle.The condenser is one of an important equipment of this circulation,which is playing the role of a cold source.The exhaust steam condenses into water in the zone of condenser,the system provides the condensate to the boiler,at the same time to form a vacuum in the condenser.The power plant condenser system is relatively complicated,the influence factors of vacuum are not easy to determine accurately,which is likely to affect the operation personnel timely and accurate processing of vacuum failure problem.Therefore,the research on safety and stability of power plant condenser fault diagnosis,environmental protection and economic operation has important significance.In the history of condenser fault diagnosis methods study normally with neural network,the fault tree analysis,expert system and fuzzy diagnosis methods is given priority to.In this paper,the research to the condenser fault diagnosis based on the idea of multiple method fusion.The main research content and method are as follows:First of all,analyze the confirmation method of condenser vacuum,find out the factors that result in the change of the condenser vacuum.Principle on the basis of consulting a large number of literature combined with power plant condenser system diagram,detailed analysis the condenser faults and the changes of parameters.In the end,the more abundant fault and fault symptom sets has been summed up.Secondly,the working principles of neural network of BP,RBF and Elman are researched by comparing different kind of neural network.In the meanwhile,the vectoring treatment example is used for the preliminary diagnosis.The results are also combined with the theory of D-S evidence in order to get the final results.Finally,considering the compatibility of the control system of power plant automation,using C# language development based on D-S evidence theory of condenser fault diagnosis system and its interface program.Through detailed study of the above content,the results show that the single neural network for the condenser fault diagnosis results are not clear,fault stripping ability is poor;the two kinds of neural network diagnosis fusion method based on D-S evidence theory,improves the accuracy of diagnostic results;instances prove that development of condenser fault system can run a good auxiliary personnel quick judgment of the accident,more accident treatment time for operation personnel,to ensure the safe operation of power plant has a certain practical significance.