A doubly fed induction generator has been widely used for extracting wind energy using variable speed wind turbines. In this paper, a developed integrated intelligent automatic control system is proposed to increase the extracting energy as well as to improve the power system performance. The controller is decomposing and grouping the sensing wind speed time series via a high-accuracy anemometer into components of reduced-order complexity using the complete ensemble empirical mode decomposition with adaptive noise and approximate entropy techniques. The output signal is applied to extreme learning machines as a real-time predictive control system. Artificial neural network, fuzzy controller, and support vector machines models are used for comparison and model validity as well as accuracy. The developed integrated PowerFactory–MATLAB simulation tool is used for obtaining, testing, and adjusting optimum wind turbine speed based on the real and predictive wind speed values in the way that increase the extracting energy by using a hybrid predictive-maximum power point tracker. The selected model that connects inputs–outputs is determined during the offline training period and, then, the predictive controller is determined online using the instantaneous inputs values. The efficiency of the developed model is verified by applying it to real wind speed data from Australia and the National Renewable Energy Laboratory. The simulation results display an excellent performance of the measured generator speed, voltage, current, active, and reactive power in addition to wind turbine power as well as a pitch angle during transient and steady-state conditions. The experimental results have verified the validity and accuracy of the developed approach and the control algorithms for practical applications.