This paper presents a new wind speed estimation method for a variable speed wind turbine system, where the theory of SVR (support vector regression) is applied. The turbine speed is controlled to capture the maximum power according to the optimal tip-speed ratio. The inputs of SVR estimator are the wind turbine power and the rotational speed. By off-line training, a specified function which relates the input with the output is obtained. Then, the wind speed is determined from this predicted off-line function and the instantaneous input. The simulation and experimental results have shown that the wind speed is estimated rapidly and accurately