Affiliation:
1. DTU Compute Technical University of Denmark Lyngby Denmark
2. ENFOR A/S Holte Denmark
3. QUENT ApS Hellerup Denmark
Abstract
SummaryWe consider reconciliation of wind power forecasts in a spatial hierarchy with three aggregation levels. We produce base forecasts for the bottom level consisting of 407 substations (connection points for local groups of wind turbines). State‐of‐the‐art forecasts from a commercial forecast provider are available for the middle and top levels, which consist of 15 regions and the entire Western Denmark (DK1), respectively. We find that the accuracy of the total forecast can be improved through spatial reconciliation, even with a relatively simple model used at the lowest level of the hierarchy. Computing the base forecasts for the substations using wind speed as the only predictor, the RMSE of the DK1 forecast is reduced by 20.5%, while the RMSE of the regional forecasts is reduced by 4.7%, on average, through reconciliation. The increase in accuracy is partly due to reduced errors in the individual regional forecasts and partly due to reduced residual correlation between the reconciled regional forecasts. We test adaptive estimation of the covariance matrix of the base forecast errors and find that it has a limited impact on the accuracy, hinting toward a time‐stable covariance structure.
Funder
ARVO Foundation for Eye Research
Subject
Renewable Energy, Sustainability and the Environment
Cited by
4 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献