Using neural network optimized by imperialist competition method and genetic algorithm to predict water productivity of a nanofluid-based solar still equipped with thermoelectric modules

The water productivity of a new nanofluid-based solar still is modeled in terms of the solar radiation, fan power, ambient temperature, glass temperature, water temperature, basin temperature, and nanoparticle concentration. The solar still is equipped with a thermoelectric cooler in which four thermoelectric cooling modules encompass the condensing channel. The Cu2O-water nanofluid is utilized in the basin of solar still. A Multi-Layer Perceptron (MLP) neural network optimized by the Imperialist Competition Algorithm (ICA) and Genetic Algorithm (GA) is employed for predicting the water productivity. The ensemble models (GA-MLP and ICA-MLP) estimate the pattern of targets better than the common MLP. Applying GA and ICA has significant effects on the accuracy of MLP, while applying ICA causes a better enhancement compared with GA. In comparison with the common MLP, the root mean square error decreases 40.49% and 62.01% in the testing phase by applying the GA and ICA algorithms, respectively. (C) 2020 Elsevier B.V. All rights reserved.