Cooling tower modeling based on machine learning approaches: Application to Zero Liquid Discharge in desalination processes

In recent years, increased emphasis has been placed on the challenge of brine disposal, driven by a heightened awareness of substantial environmental concerns associated with the utilization of desalination processes for freshwater production. By implementing Zero Liquid Discharge (ZLD) processes, desalination plants can reduce their environmental impact and improve sustainability. In a novel approach, a Multilayer Perceptron (MLP), a specific type of artificial neural network within deep learning, is used to estimate the evaporated water mass of a cooling tower system for ZLD desalination purposes, and the results are compared to traditional linear regression modeling. The TRNSYS was used to simulate the process and provide extensive evaporated mass datasets. Over 12,000 simulated data points were used to train the artificial neural network. Such network was subsequently evaluated by using real data collected from the case study. In this case, a novel example of a ZLD desalination plant installed in Cartagena-Mar Menor (Spain), including cooling tower equipment for the evaporation-precipitation process, was considered as a case study. In addition, a sensitivity analysis of the neural network model was carried out as well. From the results, it can be concluded that the MLP model has a better R2 with value of 0.76%, significantly higher than the linear model. Despite a reduction in input variables and training data, MLP excels in accurately estimating cooling tower performance, presenting advantages in computational efficiency. The sensitivity analysis reveals the model's robustness in capturing key variables, notably the significant impact of inlet brine temperature on evaporated water mass. This underscores MLP's potential for diverse scenarios and locations, facilitating cross -study applications across different environmental contexts.