Design Of Condenser Health Evaluation Software Based On Feedforward Neural Network

Nowadays,the power industry occupies a top priority in China’s energy strategy.The condenser system,as the most important auxiliary system in power plants,has a direct impact on the entire thermodynamic cycle of the power generation process.In this paper,a set of software is designed to conduct a macro assessment of the health of a condenser,give a quantitative performance index,diagnose possible faults in the condenser equipment,and provide feasible maintenance measures.Firstly,a large amount of data is collected to analyze the working principle of the condenser,the flow of the soda water and the indexes that affect its performance.The types of faults that may appear during the operation of the condenser are summarized.The consequences,causes and monitoring of the low degree of vacuum Mode,and identified the main monitoring parameters for the health assessment of the condenser.Secondly,the concept and training rules of feedforward artificial neural network are expounded.Two kinds of networks,generalized distributed neural network(GRNN)and probabilistic neural network(PNN),are selected as the main algorithms to establish the assessment model of the state of health of the condenser,which are respectively used in the state analysis and fault recognition function design.For this reason,this article compares the advantages of the two networks in terms of speed and accuracy.Thus,this paper combines the large amount of data obtained from the power plant SIS and simulation system to train the designed network to confirm its data center and weights.Then,a comprehensive requirement analysis and functional architecture design of the condenser health assessment software was carried out.According to the design idea,the software was coded and run,and the actual data was used to evaluate the performance of the software to determine its applicability.