Accelerated Continuous Conditional Random Fields for load forecasting

Hongyu Guo

Abstract: Enabling commercial buildings to use energy in a flexible way, such as reducing usage during consumption peak hours, is of crucial importance, not only for preventing the disruptions of the utility grid, but also for containing buildings' rapidly growing energy cost. Such smart energy consumption, however, heavily relies on accurate short-term energy load forecasting, such as hourly predictions for the next n (n > 2) hours. To attain sufficient accuracy, we treat such multisteps ahead regression task as a sequence labeling (regression) problem, and adopt the Continuous Conditional Random Fields (CCRF) to explicitly model these interconnected outputs. A Predictive Clustering Trees (PCT) is proposed, aiming to divide the relationships of related outputs into a set of “sub-relationships”, each providing more specific feature constraints for the interplays of the related outputs. For the multi-dimensional target space, the PCT partitions the input space, namely X, into different disjoint regions, where each is a leaf and each groups instances with similar values for the target variables Ys. Intuitively, depending on X, the PCT forms sub-relationships among targets, thus enabling the CCRF to better capture the correlations between related outputs.