Pyramid-shaped solar still desalination systems performance assessment and optimisation using sequential neural networks and genetic algorithms

• Sequential artificial neural network predicts pyramid solar stills yield with MAE 0.04 and R≈0.98. • Daytime, irradiance and temperature difference explain about 70% of productivity variance. • Optimum at depth 3.5 cm and temperature difference 11.7 °C; yield 0.720 L/m 2 .hr. • AI tool reduces assumptions and speeds PSS design optimisation. Freshwater scarcity in many regions underscores the urgent need for efficient and sustainable Freshwater production methods. This study presents a sequential artificial neural network (ANN) model, integrated with a genetic algorithm (GA) optimisation, to predict and maximise the freshwater output of a pyramid solar still (PSS). By modelling thermal and mass-transfer stages, the sequential ANN accurately captured complex nonlinear interactions among meteorological, design, and operational parameters, without relying on simplifying assumptions. The ANN model achieved high predictive accuracy (MAE = 0.04, R = 0.98), aligning well with experimental data. Applying importance analysis to the ANN process model, indicated solar irradiance, daytime, and temperature differences have the highest impact, accounting for 70% of the variance in productivity. Optimisation results indicated that the ideal operating conditions were achieved at water depths of 3.5 cm and 3.6 cm, with temperature differential of 11.7 °C for GA. These conditions yielded peak hourly productivity of 0.720 L/m 2 .hr. The optimal water depths slightly exceed the acceptable range of 1–3 cm reported in the literature, which may be due to differing control conditions in those studies. This hybrid modelling framework provides a powerful tool for improving PSS design and accelerating the commercialisation of solar desalination systems.