Productivity prediction of a spherical distiller using a machine learning model and triangulation topology aggregation optimizer

Solar stills offer a sustainable and environmentally friendly solution to water scarcity in remote areas, but their limited productivity hinders their wider adoption. This study proposes innovative modifications to the spherical solar distiller to address this challenge. We introduce a rotating spherical ball within the distiller and investigate its impact on productivity at various speeds (0-2 rpm) with and without a wick. Additionally, we explore the effectiveness of preheating feed water to different temperatures (45-70 degrees C) and its interaction with the rotating ball mechanism. Moreover, six machine learning models were employed to predict the water productivity of the distillers under different working conditions. The employed models were standalone long short-term memory (LSTM), LSTM optimized by reptile search algorithm, LSTM optimized by grey wolf optimizer, LSTM optimized by dwarf mongoose optimization algorithm, LSTM optimized by manta ray foraging optimizer, LSTM optimized by triangulation topology aggregation optimizer. The results showcased that with an optimal rotation speed of 0.5 rpm and 1 rpm for configurations with and without wick, respectively, we achieved productivity increases of 62 % and 55 %. Notably, preheating feed water to 65 degrees C further boosted the new distiller performance, surpassing the conventional solar still by 91 %, achieving an impressive output of 6000-6200 mL/m2.day compared to 3000-3250 mL/m2.day for the conventional distiller. Moreover, the thermal efficiency of the new distiller configuration reached 62 %, almost doubling that of the conventional distiller (32 %). Moreover, the triangulation topology aggregation optimizer outperformed other models in predicting water productivity with a high R2 range of 0.953-0.999.