IBM Cranks Turbine Decisions

Location is everything, especially when it comes to the placement of large, expensive wind turbines. However, choosing the right spot to plant a turbine involves a great deal of planning and data—not to mention computational might.

For companies like Denmark’s Vestas Wind Systems, decisions about turbine placement involve massive datasets. It is necessary to look at everything from seasonal wind patterns, general turbulence, and any other number of environmental and geographical factors. The company predicts that it will be analyzing even larger weather data sets in the coming four years and is anticipating sizes to reach the 20 petabyte range.

Careful consideration goes into such decisions, not only because of the high cost of turbine placement, but because once installed, these powerful energy creators are meant to stay for the long term—often for periods spanning in the decades. Additionally, those who choose to invest in the turbines want to understand the ROI in detail, thus they need to have accurate models and predictions based on highly variable datasets.

To arrive at the answers they needed to place turbines in high-value locations, Vestas Wind Systems looked to IBM. The company is using Big Blue’s IBM Big Data Analytics as well as its GPFS on the IBM System x iDataPlex supercomputer, which Vestas has knighted “Firestorm”. IBM stands its ground on the file system front, noting that for datasets that are at the scale Vestas predicts, GPFS will be capable of managing and scaling with their deluge.

According to IBM, Vestas is running the BigInsights software on 1,222 iDataPlex servers that have been strung together to create Firestorm. Running GPFS, “petabytes of structured and unstructured data are analyzed from weather reports, geospatial and sensor data, satellite images, maps and weather modeling research” to lead to the best locations for the highest wind energy ROI.

IBM claimed this week in the announcement about their environmentally-friendly customer win that as it is “capable of 150 trillion calculations per second, the Firestorm cluster replaces an HP system that was less powerful and not as energy efficient.” They  says that the same analytical process that supported previous decisions on the HP system took two weeks, but with Firestorm they’re now looking at receiving results in an hour.