Machine learning-based optimization and performance analysis of cooling towers

In the present study, a machine learning technique was employed to select parameters and optimize the performance of a cooling tower used in a building. Input parameters and the output response values have been selected in an unstructured and structured manner. Air velocity, water flow rate and nanoparticle concentration have been considered as the input variables whereas range, cooling efficiency and energy consumption are the output variables. A Central Composite design (CCD) based technique has been adopted to structure the input-output relations. A multilayer perceptron-based ANN model has been built using both unstructured and structured combinations of data points. PSO optimization has been performed by creating the objective functions from the developed ANN models. Optimum results obtained by using the structured combination are better than those obtained by the unstructured combination. Optimum cooling tower performance was noticed for an air velocity of 14.67 m/s, water flow rate of 3.87 LPM and nanoparticle volume concentration of 1.84 vol (%).