Abstract: Aiming at two common wind power forecast error estimation methods, that is, the Gaussian mixture distribution fitting method and the eigenvalue extraction method, the principle and process of these two methods are introduced in detail. Moreover, these two methods are verified by the actual wind farm data, and the estimation intervals of two methods are compared according to the impact on energy storage capacity configuration. In order to give consideration to the accuracy and validity of error estimation interval and guide the configuration of energy storage system effectively, based on Gaussian mixture model（GMM） and combined with Markov model（MM）,an MM-GMM based optimized forecast error interval estimation algorithm is proposed and verified.