Topic 9.C: Automated Classification and Validation System for Building Data

Buildings account for 30% of final global energy consumption and 28% of global energy-related CO2 emissions. Advanced analytics and controls software has been shown to curb unnecessary energy use, and generate individual building and portfolio energy savings of up to 47% and 33%, respectively. Unfortunately, the status quo to deploy these powerful technologies is slow, expensive and often inaccurate, which leads to poor return-on-investment and market adoption. Today, trained personnel are required to translate, or manually map, existing metadata from building automation systems to deploy advanced software. It may take up to 833.3 hours for a building expert to map metadata for a single software application on an average commercial building. This process is insufficient to support building energy management goals for real estate owners, operators, utilities and software vendors. This Small Business Innovation Research Phase I project will demonstrate the feasibility of a novel automated classification and validation framework to reduce the level of human effort, time, expense and inaccuracies to deploy advanced software. The framework includes four machine learning classification modules and an optical character recognition based validation model. This approach will predict probabilistic labels for 4 types of labels that are critical for deploying advanced software: equipment type, point type, equipment instance and equipment relationship. The average accuracy for each label’s prediction is anticipated to achieve at least 80%. This represents a significant reduction in time, expense and human effort required to prepare accurate data for use in advanced analytics and controls software. Validation tests and results in Phase I will inform ongoing development, field testing and integrations with existing analytics tools in Phase II. Real estate is the world’s largest asset class, yet the industry has been relatively underserved by technology when compared to consumer, medical, and financial verticals. Today, the median age of commercial buildings in the U.S. is more than thirty years old, and energy and contemporary tenant considerations are largely absent from their design and operations. New technologies and software show great potential to transform our built environment, and the rise of the millennial workforce drives the expectation of a high-tech, eco-friendly workplace. The development and commercialization of the proposed automated classification technology spurs the digital transformation of our most familiar spaces—offices, schools and hospitals—thereby increasing opportunities to improve energy and operational efficiencies, and, ultimately, improving the health of tenants and our climate.