Prediction of wind power generation based on ARMA with additive noise model

Because of a mass of factors, forecast quality of ordinary ARMA model will be greatly reduced. Aiming at the condition, the time series of wind power generation from the Fujin Wind Farm located in China and HOYW-based order dertermining method are used to build ARMA with additive noise model. Model applicability is verified by the analysis of the residual error. Applying this new model to predict, the Normalized Mean Absolute Error (NMAE) is 0.0658. Then, the superiority of the new ARMA model is verified during the comparison with the ordinary ARMA model in accuracy of wind power generation prediction.