Many improvements have been researched for conventional Savonius wind turbines in the blade shape to amend the power efficiency. In the present study, the artificial neural network was used to predict the optimum blade shape design for an enhanced power coefficient value for Savonius wind turbine at low wind speed condition numerically with commercial code software ANSYS-CFX and MATLAB. The numerical and experimental work was done with a comparison between two models; conventional and an optimized blade shape. The simulations included the analysis of many models used to learn the artificial neural network to predict the optimum blade shape of Savonius wind turbine at a wind speed of (3 m/s) and a tip speed ratio (TSR) of (0.8). Two aspect ratios (1 and 0.77) were used depending on two and three blades numerically modelled at 3 m/s and TSR range (0.2-1.2) to select best performance model for manufacturing and experimental test. From the experimental work, the torque and power coefficient were calculated based on a range of wind speed of (2.5-4 m/s) and an angular velocity range of (130-280 RPM) based on the meteorological statistics in Baghdad. The numerical results were compared with the experimental ones obtained from the wind tunnel test. The results manifested that the best model is the modified with two blades with AR= 1, and the enhancement ratio of power coefficient is 46% numerically and 31% experimentally.