Projects

Parameterization of Fundamental exchange processes from heterogeneous surfaces over arbitrarily complex terrain

Exchange processes of heat, moisture and momentum within the close proximity of the Earth’s surface can be parameterized for flat homogeneous terrain through a similarity theory. However, our understanding of these exchange processes over geometrically complex heterogeneous surfaces is incomplete. Thanks to new accelerator technologies in supercomputing and adaptive numerical methods, fundamental problems pertinent to the exchange processes over complex terrain can now be performed with fine-scale spatial resolutions on the order of meters, which enable numerical models to represent slope characteristics of complex terrain accurately (i.e. slope normal and heterogeneous surface coverage). Equally important, direct numerical simulation technology for idealized stability dependent slope flows is increasingly becoming feasible for testing hypotheses related to potential scaling laws in exchange processes. In this project our goal is to perform both DNS of idealized slope flows and LES of fine-scale complex terrain simulations to construct a scaling theory within the close proximity of heated and cooled surfaces. Data from recent extensive field experiments conducted in the U.S. and Europe will be used to validate the hypotheses and numerical simulations.

Parallel software infrastructure for adaptive simulation of Multi-scale thermal and fluid flow

High fidelity simulation of coupled thermal and fluid flow phenomena are some of the most compute-intensive applications. The simulation software is built upon sophisticated mathematical algorithms and physical models. Disparity in spatial and temporal scales in the simulation problem necessitates multi-scale adaptive methods for accurate results. The resulting software has to be parallel to simulate large problems with an acceptable turnaround time. GEM3D is a new effort to build an open-source parallel fluid dynamics simulation software. It is designed from ground up for the massively parallel architecture of modern supercomputers. The technical objective is to demonstrate scalable computations at extreme scales. Figure 4 presents an example on the adaptive mesh around a highly complex aircraft geometry. 

Data-driven modeling of an atmospheric pollutant dispersion and its event reconstruction

Gaussian plume and puff models have been formulated decades ago. These models have found many uses in applications such as air-pollution and chemical, biological and radiological agent dispersion. Gaussian models are simple empirical formulas, but they benefit tremendously when applied as part of a Bayesian inference engine driven by data collected from a sensor network. Figure 1 shows a reconstruction of passive air-borne contaminant transport field experiment. The location and the emission rate of the contaminant source is determined using a Gaussian plume model with data-driven turbulent diffusion parameters.  

WEather-driven dynamic rating of overhead power lines for transmission congestion mitigation

Congestion in existing transmission lines hinders the integration of renewable energy resources such as wind and solar. Most of the transmission network operators adopt the static line rating standard to determine how much power can be transmitted through a particular transmission lines. Static rating assumes conservative weather conditions to determine the maximum current allowed on a conductor. The primary concern is to avoid sagging of power lines to unsafe levels. Dynamic rating is a new concept that takes into account the convective cooling effect of the wind on the conductor. Weather data (wind speed and direction, temperature and moisture) close to the ground is needed to accurately initialize and update terrain-resolving micro-scale wind models. These wind models calculate the wind speed and direction at dense intervals along the path of a transmission line as it crosses over arbitrarily complex terrain. In complex terrain regions, wind prediction is highly uncertain. Reliable and dense measurements of weather close to the terrain could be used in a data-driven fashion to optimize the parameters of the micro-scale wind models and quantify the uncertainty of predictions.

Multi-scale wind modeling to forecast power production from wind farms

Wind energy is a variable resource. The utility companies rely on various methods to forecast power generation from a wind farm to balance the load on the electricity grid and also engage in energy trading and transmission scheduling. Wind operators can be penalized if they miss their target power production. PI Senocak develops a multi-scale approach where weather is downscaled to the wind park and turbine level to compute power production and potentially forecast it for the next 0-6 hours time horizon. A model-chain approach is used to address the multi-scale nature of the winds where local winds are driven by weather at the regional scale but modified with localized events and immediate terrain. The regional scale weather is also governed by the global-scale weather. Field measurements and remote sensing are an integral part of regional scale weather model initialization and updating. Figure 3 show a hypothetical wind park built over a complex terrain region. Wind turbines are parameterized as wake models and the turbulent wind field through the park and above the complex terrain is computed with a large-eddy simulation paradigm on clusters of graphics processing units in parallel.