Developing an Artificial Neural Network Algorithm Optimized for Accurate Output prediction of a Multi-step Solar Desalination System and Exergy Analysis

With global warming rates reaching record highs in recent years, the need for sustainable freshwater production is more prevalent than ever before. Modeling solar-based desalination configurations with transient simulation tools can be time-consuming, especially at larger sales. With the propagation of artificial intelligence, faster and more efficient methods can be achieved to drastically reduce simulation runtime. The goal of this paper is to assess the increased prediction efficiency and precision of artificial intelligence utilization and neural network modeling as a potential replacement for the relatively more time-consuming transient methods. The simulated solar desalination system exhibited the maximum amount of 28.2% exergetic efficiency on September 22, corresponding to 5.56 L of freshwater production. The maximum amount of freshwater production happened on March 26. By implementing the optimized ANN algorithm to obtain system output predictions, a decrease in total runtime was shown when compared to the transient simulation. The prediction results of the ANN algorithm indicated excellent correlation coefficient value of 0.99216 for freshwater production. Regarding the mean square error and root-mean-square error for freshwater production, their values were 0.016605 and 0.12886, respectively. The potential applications and benefits of the ANN forecasting methods were demonstrated in this study for a solar-based desalination system. In the future works around this topic, improvements could be made to further increase the accuracy and decrease the required time for predictive ANN analysis.