Modelling and optimization of an inverted pyramid solar still using ANFIS-PSO: predictive analysis of water production, energy, and exergy efficiency

Abstract

• ANFIS-PSO improved R 2 to 0.989 for yield and cut RMSE by 77 % compared to standard ANFIS. • 30L water volume gave 4.4 L/m 2 /day, a 64.2 % gain over 5L due to better thermal storage. • NSWL-SS raised yield by 58.3 % vs. RSS using stones, wick, and luffa for heat retention. • Energy and exergy efficiencies peaked at 55 % and 6.0 % with passive thermal materials. • PSO reduced MAE by 73 % (energy) and 62.8 % (exergy), boosting ANFIS prediction accuracy. This study addresses the pressing challenge of enhancing the predictive modelling of solar still performance, focusing on critical parameters: water yield, energy efficiency, and exergy efficiency. The research utilizes an Adaptive Neuro-Fuzzy Inference System (ANFIS) optimized with Particle Swarm Optimization (PSO) to refine the accuracy and reliability of predictions in solar desalination systems. By incorporating a modified solar still with an inverted pyramid aluminium basin and experimenting with different water volumes (5, 10, 20, and 30 litters) and natural materials such as stones, wick, and luffa, this study assesses their impact on heat retention and freshwater yield. The input variables for the ANFIS-PSO model include time, wind speed, ambient temperature, solar radiation, water quantity, and the type of natural materials used, which are crucial for understanding environmental and operational influences on solar still performance. The results show that the ANFIS-PSO model significantly outperforms the standard ANFIS model. During testing, the ANFIS-PSO model achieved R 2 values of 0.9899 for water yield, 0.9706 for energy efficiency, and 0.9642 for exergy efficiency, compared to ANFIS R 2 values of 0.8108, 0.6894, and 0.7250, respectively. Additionally, the mean error for water yield was reduced by 43 %, energy efficiency by 66 %, and exergy efficiency by 68 % in the ANFIS-PSO model, demonstrating its superior accuracy. These results highlight the potential of integrating PSO with ANFIS to enhance the predictive capability and reliability of solar desalination systems, offering valuable insights for their optimization and design.