An online optimization system was developed and applied to the condensing steam turbine network of a chemical plant. First, we developed a hybrid model of the condensing steam turbines with multiple steam injectors, by combining thermodynamic models into support vector machines. The developed hybrid model is capable of predicting the electric power generated by the steam turbines, with prediction errors of 1%-2%, and computing several performance indicators, such as the overall efficiency and the power recovery rate. An optimization problem then was formulated by utilizing the developed model to maximize the total electric power recovery from the steam turbine network. Finally, an online optimization system was developed that consists of the optimization engine (to solve the optimization problem), the model manager (to update the models), and the optimization client (to inform the turbine operators of the optimization results). The energy cost has been considerably reduced, because the optimization system was applied to the steam turbine network. The proposed hybrid modeling method can be used to predict the performance and power generation rate of various types of steam turbines in the chemical process industry.