A typical building, with a modern control system, produces about 2 million data points per day (~1500 points @ 1 per minute intervals). On the other hand, a building operator usually has very few questions that they would like answered from this data; is the building comfortable, functioning normally, and is energy wasted in the operation of the building. In a scenario were a building operator is managing multiple buildings, many of these questions can remain unanswered unless a problem arises.
In this work, we are looking at ways to quickly and easily aggregate data from building management systems to help quickly answer these questions. When information is displayed on the backdrop of a building’s architectural plan, it becomes clearly evident where portions of the building either need to be re-tuned or operated differently. Being able to perform this analysis quickly increases the quality of upkeep a building receives and can prevent operational problems by maintaining working efficiency.
|One Island East – Hong Kong, 70 story sky-scraper. The image illustrates some possible phasing issues in the control system (one zone is heating while the adjacent zone is cooling).
|Phase response of the Y2E2 building on the Stanford Campus. This method helped to tune the phasing between the atria temperatures and some of the perimeter offices in an EnergyPlus model.
Large amounts of data are often captured from either real world building sensors, or virtual building models, for many purposes including control design, fault or aging analysis, and model calibration. Because of the large dimension of this data on both spatial and temporal scales, it is often challenging to come to quick conclusions about what information of engineering importance is in the data. In this work, we present an approach to quickly assess spatial information in data based on the spectral content of the Koopman operator. We use operator theoretic methods to capture Koopman modes that represent the spatial content of oscillations in thermal quantities. By investigating these modes for different physically significant time-scales (e.g. diurnal, or control system time-scales) we can quickly capture how different areas of a building are responding to load changes at various frequencies (”breathing”).
B. Eisenhower, T. Maile, M.Fisher, and I. Mezic Decomposing Building System Data for Model Validation and Analysys using the Koopman Operator IBPSA National Conference, Simbuild NY August 2010
|The screenshot shows a selection screen of sensors installed in the Student Resources Building. Users can quickly access data by clicking on individual rooms shown on the building’s floor plan.
|Magnitude of the Koopman mode for the 24 hour oscillation over the floorplan of a building. Colors signify the magnitude of spectral content for the 24-hour period shown. Red areas contain high spectral content while blue areas contain little spectral content.
|Phase of the Koopman mode for the 24 hour oscillation. Phase is relative to the temperature of outdoor air. Red areas correspond to regions whose temperature behavior is in phase with the temperature of the environment while blue areas correspond to regions whose temperature behavior lags that of the environment