Regresar

Reconciling solar forecasts: Sequential reconciliation

Abstract:

When forecasting hierarchical photovoltaic (PV) power generation in a region and/or over several forecast horizons, reconciliation is needed to ensure the lower-level forecasts add up exactly to the upper-level forecasts. Previously in “Reconciling solar forecasts: Geographical hierarchy” [Sol. Energy 146 (2017) 276–286] and “Reconciling solar forecasts: Temporal hierarchy” [Sol. Energy 158 (2017) 332–346], forecast reconciliation has been demonstrated for geographical and temporal hierarchies, separately. This article follows such frameworks and extends the reconciliation to spatio-temporal cases. More specifically, sequential reconciliation is used for operational day-ahead forecasting of 318 PV systems in California. It is shown that by using sequential reconciliation, aggregate-consistent forecasts can be obtained across both the geographical and temporal hierarchies. In addition, the forecast accuracy can be further improved from that of the single-hierarchy cases.