Comparative analysis and application of cooling tower outlet water temperature prediction methods

Aiming at the defects of the traditional BP neural network prediction method, proposes a genetic algorithm optimized BP neural network prediction method. Based on the measured data, compares and analyses the three prediction methods of enthalpy difference method, traditional BP neural network method and optimized BP neural network method, and analyses the effects of circulating water flow, inlet water temperature, wet bulb temperature and dry bulb temperature on the cooling tower outlet water temperature. The results show that the mean square error of the tower outlet water temperature obtained by the optimized BP neural network method is 0.000 787 ℃2, the average relative error is 0.079 9%, and the root mean square error is 0.028 069 ℃, which is significantly smaller than other prediction methods. Using the improved prediction model to analyse the annual operating data of a company’s open cooling tower, it is found that the outlet water temperature is positively correlated with the circulating water flow, the inlet water temperature and the wet bulb temperature, and negatively correlated with the dry bulb temperature.