Wind Speed Modeling Based On The Extended Kalman Filter

In order to improve the performance of wind power systems, we need establish precise wind speed models. Based on the Von Der Hoven spectrum, a wind speed profile is divided into the low frequency wind speed and the high frequency one. This thesis considers a parameter estimation problem of a wind speed model for the low frequency component.To simulate wind speeds, we adopt a traditional low frequency wind speed model in terms of the sum of a number of cosine signals. The state space equation is established by using the amplitude, frequency and phase parameters of cosine signals as state variables. We first consider the parameter estimation problem of a single cosine signal by using real and complex Kalman filters respectively. We then extend this idea to use the combination of real and complex Kalman filters as parameter estimators. For the case of two cosine signals, we also present a similar estimation algorithm. The algorithm is subsequently used to establish a low frequency wind speed model for a wind speed data. Simulation results show that the wind speed model obtained by the two-cosine-signals algorithm can approximate the low frequency component of the wind speed.