Study on soft-sensing model for condenser vacuum based-on Support Vector Regression

In this paper, depending on the interrelation of condenser´s operational parameters, the factors which affect the vacuum of condenser are analyzed. And a soft-sensing model for condenser vacuum is given by using Support Vector Regression (SVR), then the model is verified and parameters are discussed based on the data of the 300MW steam turbine unit, and the prognostication precision is compared with a RBF model. The results indicate that model based-on SVR has forcible generalization ability and stability and can be adapted to application. The condenser vacuum can be calculated by using the soft-sensing model when the vacuum measuring point is fault, so the model based-on SVR provides a redundancy method for the measurement and diagnosis of condenser vacuum.