Calculating the impact of meteorological parameters on pyramid solar still yield using machine learning algorithms

Deterimining the effect of individual meteorological and solar still parameters on solar still productivity is challenging. This is because the parameters are correlated and experimentally isolating their individual impact is difficult. However, locations of the solar still that optimize productivity can be determined by analyzing the effect of these parameters on yield. In this study, two algorithms, an artificial neural network (ANN) and linear regression (LR), were used to predict the yield of solar still. Four important parameters were identified by analyzing the effect of different parameters on the yield and used in prediction models. The number of input features was reduced using principal component analysis (PCA) and a correlation function, which reduced the complexity of the ANN model. Based on correlation and feature importance, the features of Solar, Tf, Tout, and Wind had the most influence on productivity, as demonstrated by the accuracy and model complexity. ANN and LR were trained to identify models that fit to the data with minimum error. The ANN model perfomed better than the LR model, as demonstrated by the mean absolute error (MAE), R-squared (R²), and root mean square error (RMSE), because it can model nonlinear systems. The performance of a solar still is influenced by various meteorological parameters; however, determining the specific effect of each parameter can be challenging with a low-complexity ANN. By excluding parameters that have a weak effect on the yield and eliminating those with strong dependencies, four parameters that accurately predict the solar still's yield were identified. This approach reduced the ANN complexity by 72% compared to using all 12 general parameters. The identified parameters can be used by researchers who conduct several days of experiments that determine solar still yield; they can compare their findings for different values of these parameters. In addition, the location of solar still can be determined based on the importance of these parameters.