Soft Sensor Modeling for the Efficiency of Steam Turbine Last Stage Group Using Support Vector Machine Regression

To calculate the steam turbine exhaust enthalpy, this paper proposes a soft sensor method by using the support vector machine regression (SVR). The proposed method is based on the following three-step strategy. Firstly, main factors, influencing on the last stage group efficiency, were discovered through mechanism analysis. Secondly, based on the designed sample data, the support vector machine regression is used to establish the functional relationship between the exhaust enthalpy and these main factors. To identify the parameters involved in the SVR, the genetic algorithm (GA) is taken as the optimizer. Finally, some experimental sample data collected from a 600MW unit are used to validate the established soft sensor model. The results show that the proposed method has high prediction accuracy, by comparing with thermal test data.