Pattern analysis, Grid incentives, Building thermal energy storage, Intelligent building controls, Building performance optimization
Building thermal energy storage has been utilized for decades for various objectives, such as reducing peak electrical demand, reducing building operating expenses, and increasing the efficiency of systems when charged from waste heat or free cooling. As building thermal storage control strategies become more dynamic, optimization of building performance often considers multiple objectives that aim to improve building performance in energy, economic, environmental, and grid support categories. The dynamics of the incentive signal used for one objective, as well as its relation to signals from other objectives—for instance, whether the signals are “in sync” or are “conflicting”—heavily influence the tradeoffs that may exist among performance objectives. To better understand the degree of alignment that may exist between grid incentive signals, we apply unsupervised learning to a novel grid data set that includes hourly signals for energy price and marginal carbon emissions. Clustering algorithms identify common patterns in the dynamic signals. Overall, Hierarchical Clustering demonstrated the best performance, evaluated by DB index and Silhouette score. While the algorithms did not find distinctive patterns among the carbon signals, they did identify 7 to 9 patterns within the January and July pricing signals. The highly fluctuating nature of the carbon emission signals could lead to a diverse range of tradeoffs between building energy cost and carbon emission reduction objectives, if the signals were used as the basis for a building control optimization problem. This finding iterates the importance of understanding incentive signal dynamics, in both individual and collective contexts, and the implications for development of new control technologies for grid-interactive buildings.