考虑多维气象数据与时间影响的风电功率区间预测OA

Existing wind power interval forecasting models based on forecasting errors typically consider only the in-fluence of wind speed while neglecting the temporal effects in the modeling process.To address these challenges,this paper proposes a wind power interval forecasting model that combines the point forecasting model of neural hier-archical interpolation for time series forecasting(N-HiTS)with the theory of generalized additive model(GAM)-D-Vine copula based quantile regression(DVQR).First,the N-HiTS is employed to obtain point predictions of wind power and corresponding forecasting errors.Then,a DVQR model for the forecasting errors is constructed,where temporal variables are incorporated via P-splines GAM to model the correlation coefficients corresponding to Copula parameters,thereby obtaining conditional quantiles of the forecasting errors.Finally,the point predictions are cor-rected based on the conditional median of errors,and forecasting intervals are generated by superimposing the condi-tional quantiles of errors.Validation using actual operational data from a wind farm in Shanxi Province,China dem-onstrates the effectiveness and superiority of the proposed method.