Mechanical ventilation performance assessment in several office buildings by means of Big Data techniques

Mechanical ventilation performance is a key issue related both to energy efficiency and indoor air quality. There are several techniques for measuring ventilation rates in buildings, such as blower boor tests, flow hoods, VAV box measurements and tracer-gas techniques. From several decades ago, tracer-gas techniques are recognized as the most widely employed method to estimate air exchange rate in buildings. These methods are based on the study of the temporal evolution of the concentration of an injected gas. These methods are usually expensive and do not allow the occupancy of the building during the tests. This issue also limits severely the number of buildings to be evaluated. Despite the natural presence of CO2 in the atmosphere, there has been a growing interest in using it as a tracer gas. It has been shown recently that it is possible to estimate the ventilation rates by means of the decay method using in-situ CO2 measurements from transmitters available on the market.
On the other hand, there is a growing tendency to move to the smart paradigm, which usually implies to add more and more sensors to the buildings to achieve better controls, among other applications. Modern Building Energy Management Systems (BEMS) allow the access to a high quantity of data in different formats and resolutions, such as signals from sensors and actuators, set-point temperatures, schedules, digital switches, alarms, plots or reports.
In this work in-situ measurements are combined with data coming from BEMS to evaluate the mechanical ventilation performance of three different office buildings in different regions of Spain. Starting from metabolic CO2 production from occupants of the buildings, a series of conditions are identified in to be able to evaluate the mechanical ventilation performance. Those conditions are translated into an algorithm, programmed in the Python language, which access the different sources of information to wrangle, cluster and finally calculate the mechanical ventilation performance. This operation is performed for five years of available experimental data and information in different formats, performing a search in more than 100 Gb. of information. This situation falls into the computational framework known as Big Data, as stated by the ICT community.
For the first of the buildings (located at Almería) the methodology is detailed and demonstrated through the complete series of five years of data. Thanks to the application of this technique, a mechanical ventilation rate is obtained in perfect agreement with the design values in different situations. The technique is then applied in two different buildings (located at Madrid and at Valladolid) to assess their respective ventilation rates.
Finally, some conclusions are summarised and possible improvements of the method are pointed out as future work.