Online Reconciliation Method for Short-Term Hierarchical Load Forecasting

Hierarchical load forecasting (HLF) is a type of multi-load forecasting problem with cumulative consistency, which is usually decomposed into multiple single load forecasting and hierarchical optimization problems. However, the prediction results of this model are often biased. In addition, compared with offline learning methods, online learning methods are considered to be a more effective and adaptive prediction method. In this paper, we propose an improved reconciliation method on batch and online model that guarantees unbiased prediction results. Case studies on the power load data of one province in China shows that this method is superior to traditional methods in terms of prediction accuracy and adaptability.