Comparative study and multi-objective optimization of solar stills using priority-based clustering and prediction modelling

Abstract

The rising scarcity of potable water, coupled with environmental concerns over conventional purification methods, has accelerated the global shift toward renewable energy-based clean water solutions. Solar stills, with their simplicity, sustainability, and low operating costs, present a promising option for decentralized water purification. This study examines three solar still designs—rectangular, conical, and hemispherical—under identical conditions to identify the most efficient configuration based on key performance outcomes. However, the intricate nonlinear relationships among the output parameters and the geometric complexities of the stills pose significant challenges to conventional optimization techniques. To overcome these limitations, a hybrid methodology is employed, incorporating a priority-based clustering model to systematically optimize and identify the ideal solar still design. The integrated analysis reveals strong thermal-performance relationships, with heat transfer coefficient (h) showing high correlations with Tw, Tg, and Ta (r > 0.77, Adj R2 ≈ 0.89), while Qew is strongly linked to Tg (r = 0.9284), Tw (r = 0.8941), and h (r = 0.9221); Mew exhibits an almost perfect correlation with Qew (r = 0.9998, Adj R2 ≈ 0.9971). The ANN model using the Gaussian membership function demonstrated the highest prediction accuracy with R2 = 0.997, RMSE ≈ 0.14%, and MSE ≈ 0.01%. Priority weighting favoured Mew at 39%, followed by η (28%), Qew (19%), and h (14%), improving clustering accuracy. K-means clustering pinpointed Trial 27 as optimal, with I(t) = 820 W/m2, Ta = 37 °C, Tw = 59 °C, and Tg = 53 °C, yielding the best thermal synergy for maximum solar still performance.