In the context of achieving “dual-carbon” goals

power units function as adaptable power sources for integrating new energy sources

posing significant challenges to their power generation flexibility. The dry-wet joint cooling system plays a pivotal role in ensuring the safe and stable operation of power units. Therefore

there is an urgent need to optimize the operational strategy of the dry-wet joint cooling system to enhance its flexibility and economic efficiency. Focusing on the dry-wet joint cooling system of a 660 MW generator set

a multi-layer perceptron(MLP) neural network model has been established to predict the outlet temperature of the cooling water. A mixed integer nonlinear programming(MINLP) model is formulated and linearized based on actual operating condition constraints. By solving the MLP-MINLP optimization model

the optimal operation strategy for variable-frequency fans in each operating condition of dry-wet joint cooling system is determined

successfully reducing its power consumption. The results indicate that

after optimizing the configuration of variable-frequency fans

there is a significant reduction in total power by approximately 11.16%

and implementing different frequency operations for variable-frequency fans can reduce total power by about 3.62%～5.38% in a limited manner. The MLP-MINLP optimization model can achieve precise and low-power operation of dry-wet joint cooling system

offering a viable solution for optimizing dry-wet joint cooling systems.