Challenges and Opportunities of #HVAC Data – Gabriel from Ecorithm
Posted on June 24th, 2015
NYC smart city and energy data #smartcitydata
06/23/2015 @Urban Futures Lab, 15 MetroTech, 19th Floor, NY
Gabriel Peschiera @ Ecorithm (a software startup that turns building data into actionable insights) spoke about the challenges of collecting and analyzing sensor data within large buildings. These analyses have the potential to improve the comfort of the occupants and the functioning of the building systems.
Ecorithm focuses on buildings with more than 100,000 sq ft of floor space. These buildings have a building maintenance system to monitor the function of the chiller -> air handlers -> Variable Air Volume
In older systems, they will work with 3rd parties to place sensors for input into JACE, the Java Application Control Engine. For newer systems, they will collect data from the current BMS. In a typical building, their system receives information from 3000 locations, with sensor readings every 5 minutes. Data are available at the end of each day.
Gabriel touched on some of the data analysis tools and challenges
- Data cleaning problems often center around mislabeled locations and types of sensor inputs
- Spectral analysis in frequency space shows the temperature fluctuations driven by daily schedules, control loops and weather. From frequency plots they can see periodic patterns that might indicate persistent problems. They can also see if disparate locations have similar frequency patterns which may be driven by a common source.
- Support vector regressions can detect nonlinear deviations from expected patterns. This method can be used to detect faults and fill in missing data.
- Model-based optimization may eventually be a technology to better tune systems.
He also presented screen shots of their monitoring control panel and fault reports.