Abstract: In order to prevent the condenser scaling from causing fire and explosion accidents, it is necessary to predict the law of development of fouling factor in the condenser. A fouling factor predicting model combining K-mean algorithm and Chebyshev neural network was designed. Aiming at the disadvantages of Chebyshev neural network, the K-mean algorithm can be used to improve the curve of fouling factor development over time, which can be divided into three stages: starting stage, adhesion stage and aging stage. Results showed that, the modified Chebyshev neural networks can predict the law of development of condenser fouling factor effectively, and is more accuracy than progressive prediction and power-law prediction; the algorithm is simple and has fast convergence speed.

Key words: coal power station, fire safety, condenser, neural network, fouling factor