Prediction of Mechanical Power of New Design of Savonius Wind Turbine Using Various Empirical Models

In this paper, the Multi-layer perceptron neural network (MLPNN) and Radial basis function neural network (RBFNN) have been used to predict the mechanical power (MP) of the new design of the Savonius wind turbine. Moreover, the accuracy of the proposed models is compared with the ARIMA model and Multiple Linear Regression (MLR). In this study, the empirical models were developed to predict the mechanical power based on the various design parameters, mechanical torque, and angular rotational. The mechanical torque and angular rotational of 144 rotors were measured experimentally with various wind velocities. The results demonstrated MRIMA and MLPNN models have the best performance compared to RBFNN and MLR for MP prediction. Among the developed models, the MLPNN is presented as the best model for the mechanical power of the proposed rotors.