|Abstract Title:||Energy efficient high-accuracy urban flow predictions using AI-aided Computational Fluid Dynamics; dynamically updated catalogue of non-stationary emission sources and IoT sensor networks|
|Presenter Name:||Mr Francisco Ramirez-Javeg|
Abstract Information :
According to the United Nations estimates1, , 68% of the global population is expected to live in large urban areas by 2050. This poses a significant challenge for urban planners and decision-makers since air quality in urban areas is mostly conditioned by urban micrometeorology and non-stationary emission sources, presenting strong spatial gradients within urban geometries, e.g. changing between opposite sides of a street. Understanding the behavior of these changes is essential for tracking human exposure in urban areas and the correct design of urban spaces and mitigation measures.
The typical approach to provide high resolution modelling in urban areas, e.g. ADMS Urban, is to combine the rough snapshots provided by CAMS with air quality modeling techniques, such as Computational Fluid Dynamics (CFD), on-site Air Quality Monitoring (AQM) technologies and emissions catalogues. This enables remote image resolution upscaling and provides real-time air quality solutions with high resolution (~m2) at ground level. Nevertheless, this strategy presents some market penetration barriers i) low-cost air quality monitoring usually lacks data reliability and understandability and ii) high-fidelity CFD simulations have high computational cost and long run times, which makes it an unsuitable or unaffordable tool for providing nowcasting information or real-time predictions in several scenarios.
Bettair Cities and the Barcelona Supercomputing Center (BSC) have been collaborating over the last years to developed an alternative solution which includes the Bettair cloud platform (data analytics, GIS, and alarms for end-users) and the Bettair Air Quality Mapping (BAQM) service that can produces 2D air quality data with a resolution of ~10m2 at ground level. Nevertheless, the current model is based on simplified physical models with strong physical assumptions. To drastically improve the physical foundations and, thus, the accuracy and reliability of Bettair’s solution, Bettair, and the BSC recently develop the state-of-the-art CFD emulator (CFDE), which uses Artificial Intelligence to produce high-fidelity CFD-like results but at a fraction of the typical CFD computational cost. The emulator mimics the CFD simulation results but efficiently using the energy and computational resources. The CFDE was validated versus experimental results collected in a pilot through 2021-2022 in El Prat del Llobregat.