A novel two‐stage interval prediction method based on minimal gated memory network for clustered wind power forecasting