Solar Stills (SSs) are an eco-friendly and efficient approach to generating drinking water from brackish or saline sources. In this paper, a novel model for predicting the productivity of Hemispherical SS is proposed. This model enhances traditional MLP using an advanced Genetic Algorithm (GA), which mimics the process of “seeded selection” in natural selection to determine optimal MLP parameters. The resulting model, named Enhanced Genetic Algorithm – Multi-Layer Perceptron (EGA-MLP), is compared against two other models, standalone MLP and Particle Swarm Optimization with MLP (PSO-MLP), as well as experimental results from SSs. The models were applied to predict the yield of three different SS designs, including Conventional Hemispherical Still Without Fins (CHSWF), Cylindrical Finned Hemispherical Still (CyFHS), and Conical Finned Hemispherical Still (CoFHS), under varying environmental conditions and fins spacing. Statistical criteria such as Root Mean Square Error (RMSE), coefficient of determination (R2), and Mean Absolute Error (MAE) were used to assess the models' performance. The EGA-MLP model exhibits the highest R2 values across all tests, with an average R2 score of 0.994. In comparison, the PSO-MLP and MLP models achieve average R2 scores of 0.949 and 0.93, respectively. Furthermore, the EGA-MLP model demonstrates the lowest values of Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) among the three models. The proposed model demonstrated the most accurate predictions of SS yield compared to both the experimental results and the other two models in predicting SS yield and has the potential to improve the design and optimization of SS systems.
|
