Data-driven modeling approach for performance analysis and optimal operation of a cooling tower

This paper proposes a data-driven adaptive modeling approach to investigate the performance and optimal operation of a cooling tower for energy conservation. To achieve this aim, the cooling tower process was first characterized by an adaptive model with nonnegative garrote (NNG) variable selection procedure, which ensured a compact and robust input-output relation. Owing to the high accuracy of the obtained model, implementing the optimal operation strategy for energy saving became readily practicable. Subsequently, on the basis of the statistical results of NNG variable selection, the effects of ambient air temperature and humidity on the cooling capacity of the tower were investigated by principal component analysis (PCA). Finally, the optimal strategy of fan operation was proposed and its implementation was virtually studied based on data from the actual operation of a cooling tower, which showed that there was considerable room for energy conservation. This is the first attempt to use the NNG variable selection method for developing model for cooling tower and to propose a model-based control scheme for operating a cooling tower.