Apex Clusters Increase Efficiency and Save Money!
The mission of MIT is to advance knowledge and educate students in science, technology and other areas of scholarship that will best serve the nation and the world in the 21st century — whether the focus is cancer, energy, economics or literature. With 78 Nobel Laureates, 53 National Medal of Science winners among current and former faculty, MIT is recognized as one of the top learning and teaching institutes in the world. For more information, visit their website at www.mit.edu.
Solar photovoltaic (PV) panels are commonly assembled in a two-dimensional (2D) form, while plants and other sunlight collecting structures in nature are three-dimensional (3D). Using 3D rather than 2D can allow one to increase the generated energy per base area, at the cost of higher solar cell material per unit of generated energy. One major barrier to widespread use of solar panels is installation cost (rather than active material costs), and so by studying different ways to configure them, they also set out to explore new paradigms for solar photovoltaic installation.
Professors Jeff Grossman and Marco Bernardi proposed to:
- Write a computer code that would enable calculation of generated energy by an arbitrary 3D photovoltaic (3D-PV) structure, at any latitude and longitude and using panels of arbitrary reflectivity and efficiency, and using an accurate apparent sun trajectory. The code allows one to calculate the energy generated over a period of time from a fixed structure, or alternatively evolve the structure with an optimization algorithm to maximize the generated energy. They implemented both genetic algorithm and a Monte Carlo optimization schemes.
- Study the 3D shapes with higher energy per footprint area unit, and study the power generation profile during the day; compare these figures with flat PV panels
- Study these structures at different latitudes and during different season
- Validate their results with outdoor experiments on real 3D-PV structures
- Eventually, include weather models and wavelength-dependent effects, and parallelize the code.
Why a Cluster?
Prior to their cluster purchase, MIT was running these simulations on a laptop. While the code is not parallelized in its current implementation, genetic algorithm and Monte Carlo optimizations can be very demanding, with the longest performed calculations taking up to one month on a single processor, which is not feasible on a laptop.
So they looked for a more efficient way of running their simulations, which is where Advanced Clustering Technologies entered the picture. MIT’s first cluster purchase from them gave them the possibility to run simulations on many processors in parallel and to carry out very large-scale simulations. The code in its present form scales like N*logN, where N is the number of panels in the simulation. Currently, they are investigating entire residential areas where houses and buildings are coated with solar panels, thus forming huge 3D-PV structures with more than 10,000 panels. This clearly requires a cluster to run in a reliable way. And the one provided by Advanced Clustering delivered a decrease in run time by a factor of 100, which allowed for more simulations and a better understanding of what works and what does not.
Why Advanced Clustering Technologies?
The cluster provided by Advanced Clustering Technologies offered an excellent trade-off between cost and performance, together with competent technical assistance from the seller. This has boosted MIT’s capability to carry out their solar panel project to find a better way to deliver solar energy to the home.
Fig. 1 (a) 3DPV structures made using Si solar cells with area 3×3 cm2. From left to right, an open cube (1), an open parallelepiped twice as tall (2), and a tower (3). The structures are made up, respectively, of 9, 17, and 32 solar cells. (b) Power generated by a flat Si panel at various tilt angles measured under simulated solar light illumination, and comparison with computer simulation. The error bars in the simulation results derive from a range in the assumed efficiency of +/-1%. (c) Both measured and simulated power during a single sunny day for the open cube and for a flat panel of the same base area, showing a maximal range of hours of constant power generation and nearly twice the energy density output for the 3DPV case compared to the flat panel. (d) Energy generated by the structures shown in (a) under different weather conditions, expressed as a ratio to the energy generated by a flat panel under the same weather conditions. Comparison of the black and blue bars for the case of the parallel-piped and tower shows how structures of higher aspect ratio than the open cube can further outperform a flat panel on a cloudy day compared to a clear day. The parallelepiped in (a) is referred here as ‘‘tall cube’’. (e) Power generated vs. time for the data of cloudy weather shown in (d).
The cluster offered a reliable and efficient platform to perform calculations and back up results. At the same time, jobs can run in background while previous calculations were analyzed. Given the large amount of data generated, the 3D-PV project would have been impossible without this strategy.
MIT’s simulations demonstrated that the performance of 3D-PV structures scales linearly with height, leading to volumetric energy conversion, and provides power fairly evenly throughout the day. Furthermore, they showed that optimal 3D structures are not simple box-like shapes, and that design attributes such as reflectivity could be optimized using three-dimensionality.
This work resulted in two papers and is also featured in several websites – for example, see: http://web.mit.edu/newsoffice/2012/three-dimensional-solar-energy-0327.html
In summary, the cluster provided by Advanced Clustering Technologies facilitated the prediction of generated energy and optimal shapes, and is an indispensable tool for optimizing solar energy generation. MIT’s results show that 3D sunlight collection has the potential to serve as a paradigm shift in solar energy conversion toward the Terawatt scale.
- Case Study: Weather Forecasting in the Cloud
- Building an HPC Solution with Flexibility to Meet Future Needs
- Cost and Performance Enhance Research Efforts
- Closing the Gap On Memory Performance For Development Of Genetic Data Analysis Software
- The Virtual Campus: One of the Nation’s Largest Passive Stereoscopic Tiled Wall Displays
- Apex Clusters Increase Efficiency and Save Money!
- University Adapts to Computational Computing with eQUEUE
- Cluster Speeds Molecular Dynamics Research with NVIDIA GPUs
- Enabling Computational Research for the ‘Missing Middle’