The heat transfer characteristics of condensers and the influence of vacuum degree on unit operating economy were investigated.The implementation details of input layer

hidden layer and output layer of the clustering radial basis function(RBF)neural network were introduced

and each part of the matrix matching requirements was presented.The particle swarm optimization(PSO)algorithm was applied to look for the radius and output layer weights of the RBF model

and a concrete implementation process was put forward.The condenser vacuum soft measurement model was established.Moreover

taking the practical running data of a 300 MW unit as an example

data training for the above model was carried out.Comparison between the predicted steam condenser vacuum value and the monitoring data verifies this established model can judge the condenser operation state reasonablely.