The optimization of circulating water system in a thermal power plant holds a great significance for determining the optimum vacuum degree of a condenser and improving total efficiency of the plant to save energy. Therefore, a prediction model for steam turbine output power is established based on least squares support vector machine （LSSVM） by considering the circulating water system of two 600 MW steam turbine units in a particular thermal power plant. There is no local minimum in the model and excellent prediction results can be achieved for a variety of problems. Then, an optimization model for the vacuum degree in the condenser is developed based on maximization of the profit by ta- king circulating pump shaft power, turbine power increment, the price difference of coal and electricity on the market into account. Initially, the recorded operational parameters of the plant over a period of time are input into the model, a data pre-processing is conducted on the parameters and their stability is identified. Then the vacuum degree is optimized phy-based optimization algorithms （ SA – BBO） to by means of the simulated annealing and biogeogra- obtain the optimum vacuum degree, best combination of the operational parameters of the condenser and circulating pump under various operating conditions. The optimized results have been made into an optimized combinations chart to guide the operaion and regulation of steam turbines.