Artificial neural network-assisted modeling of thermal performance in hybrid nanofluid-integrated solar water pumps

In recent years, artificial intelligence techniques, particularly artificial neural networks (ANN), have emerged as powerful tools for accurately modeling nonlinear systems, accelerating computational analysis, and optimizing energy-related processes. This study presents an ANN-based computational framework for the thermal performance optimization of hybrid nanofluid-powered solar water pumping systems utilizing Parabolic Trough Solar Collectors (PTSCs). The Levenberg–Marquardt backpropagation algorithm is employed to train and validate the ANN model, enabling precise prediction of thermal and flow characteristics under varying operating conditions. The working fluid is formulated by dispersing molybdenum disulfide (MoS 2 ) and zinc oxide (ZnO) nanoparticles in kerosene oil, enhancing thermal conductivity and heat transfer rates. A coupled dual-phase flow model, incorporating porous media effects, thermal radiation, viscous dissipation, and nanoparticle shape factors, is developed and reduced via similarity transformations. The governing nonlinear partial differential equations are numerically solved using a shooting method in conjunction with the MATLAB bvp4c solver to generate training and testing datasets for the ANN. Parametric studies demonstrate that increasing porosity reduces velocity in both phases, while higher fluid parameter values enhance velocity gradients but lower dust-phase temperatures. Moreover, stronger fluid–particle thermal interactions decrease the fluid-phase temperature while increasing the dust-phase temperature. The proposed ANN-based framework offers an efficient predictive and optimization tool for improving the thermal performance of solar water pumping systems, supporting the advancement of sustainable and high-efficiency renewable energy technologies.