Reconciling temporal hierarchies of wind power production with forecast-dependent variance structures

Abstract

For electricity grid operators, the planning of grid operations depends on having accurate models for forecasting wind power production on a number of different time resolutions. Together, these time resolutions create what has become known as a temporal hierarchy. Previous studies have considered methods of reconciling forecast hierarchies inspired by least squares methodologies that produce coherent and more accurate forecasts. In this study, we highlight some challenges in the established approach when applied to wind power production, and consider methods which more appropriately take into account the full conditional probability densities. We suggest methods using maximum likelihood techniques to estimate the prediction variance ahead of time. Using base forecasts from a commercial forecast provider together with simpler forecasting models, we test the modified approach against the established reconciliation approach on data from Danish wind farms. The results show significant improvements in accuracy when compared to both the state-of-the-art commercial forecasts and the simpler models.