Apples or oranges? Identification of fundamental load shape profiles for benchmarking buildings using a large and diverse dataset. (15th February 2019)

• Record Type: Journal Article
• Authors: Park, June Young; Yang, Xiya; Miller, Clayton; Arjunan, Pandarasamy; Nagy, Zoltan
• Is Part Of: Applied energy. Volume 236(2019)
• Page Start: 1280
• Page End: 1295
• Publication Date: 2019-02-15
• Subjects: Building energy -- Load profile -- Energy benchmarking -- Unsupervised learning -- Data analytic -- Visual analytic; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2018.12.025
• Languages: English
• ISSNs: 0306-2619

Abstract

Graphical abstract: Highlights: Three fundamental load shape (FLS) profiles exist for building energy use. FLS profiles are characterized by morning, mid-day and evening peaks of energy use. Benchmarking using FLS profiles results in homogenous groups regardless of type. Results based on large and diverse dataset: 3829 buildings and 75 use programs. Case study shows application of FLS for building portfolio management. Abstract: Buildings are responsible for 30–40% of the anthropogenic greenhouse gas emissions and energy consumption worldwide. Thus, reducing the overall energy use and associated emissions in buildings is crucial for meeting sustainability goals for the future. In recent years, smart energy meters have been deployed to enable monitoring of energy use data with hourly or sub-hourly temporal resolution. The concurrent rise of information technologies and data analytics enabled the development of novel applications such as customer segmentation, load profiling, demand response, energy forecasting and anomaly detection. In this paper, we address load profiling and benchmarking, i.e., determining peer groups for buildings. Traditionally, static characteristics, e.g., primary space use (PSU) together with the annual energy-use-intensity (EUI) have been used to compare the performance of buildings. Data-driven benchmarking approaches have begun to also consider the shape of the load profiles as a means for comparison. In this work, we identify three fundamental load shape profiles that characterize the temporal energy use in any building. We obtain this result by collecting a dataset of unprecedented variety in size (3829 buildings) and primary use (75 programs), and applying a rigorous clustering analysis followed by entropy calculation for each building. The existence of fundamental load shape profiles challenges the man-made, artificial classification of buildings. We demonstrate in a benchmarking application that the resulting data-driven groups are more homogeneous, and therefore more suitable for comparisons between buildings. Our findings have potential implications for portfolio management, building and urban energy simulations, demand response and renewable energy integration in buildings.