This study proposes an artificial neural network (ANN) method to predict the performance of a Savonius wind turbine based on existing data. Using computational fluid dynamics (CFD) simulations with ANSYS Fluent software, simulations are conducted to gather data to train the ANN model to predict rotor performance. First, the effects of several parameters, such as the gap ratio s* and blade thickness ratio, are analyzed through CFD simulations. These parameters are then used as input data to predict performance. For the training and testing processes, 77 numerical cases are used, with 80% of the samples used for training, 10% for testing, and 10% for validation. The network includes a hidden layer; an input layer consisting of blade thickness t, tip speed ratio (TSR), free stream velocity V, rotor diameter d, and gap ratio s*; and an output is the power coefficient Cp. After training, this method is applied to predict the results for the s* 0.1 rotor blade, with the power coefficient curve showing a trend similar to the numerical simulation results, with a maximum prediction error of 9.38% at a TSR of 1.3. These results indicate that this is a reliable method that significantly reduces computation time, allowing for the evaluation of many new blade designs with similar parameters and optimizing profiles to improve the performance of Savonius vertical-axis wind turbines in the future.