In order to improve the efficiency of the cogeneration system which integrates turbine generator (TG) and cooling tower (CT), a real-time optimum operation strategy of fans is proposed. First the statistical models of TG and CT are developed off-line by using the local model network (LMN) algorithm. Then the optimal outlet temperature of cooling water (T-cw,T-out) is calculated by solving the optimization problem which maximizes the net power output of cogeneration system. Based on the calculated T-cw,T-out, a statistical linear model is employed, which characterizes Approach of CT. Finally, based on the proposed Approach model, an optimum operation mode table for the six fans is established. In order to decide optimum mode for fans, the factors such as different climatic conditions are also incorporated. Using the optimum operation table, a real-time operation mode of fans can be achieved. The performance of the proposed method is similar to the previously developed LMN method (about 85-95% as demonstrated in Table 3) while requiring a very low computational cost. The proposed method is also advantageous because the field operators can understand the physical meaning of operation. This algorithm can be easily implemented into the existing distributed control system making it a very good option for on-line implementation. (C) 2013 Elsevier Ltd. All rights reserved.