Colorado, US — Selecting the best geographic locations to build wind farms that will produce the maximum energy with the least intermittence in power generation is the perfect problem for a supercomputer to solve. And a new business model may now offer cloud computer capabilities even to small developers.
‘The supercomputer is able to take the numerous variables that will impact the proposed wind farm’s efficiency – topography, land cover, historic wind data, global climate change, proximity to transmission infrastructure – and quickly model them in billions of different combinations to pick the optimal sites,’ said Earl J. Dodd of the Rocky Mountain Supercomputing Center (RMSC).
He is quick to add that while supercomputing technology, also referred to as high performance computing (HPC), is well known for its processing speed, people often forget it offers enormous processing capacity. And this may be more important to the wind energy business as scenarios can be modelled on an increasingly smaller grid size, making it possible to pinpoint optimal sites.
‘Accurate site selection can reduce the financing and operating costs of wind energy development projects, which ultimately allows lower per-kilowatt rates for the consumers,’ said Dodd. He noted that once a wind power facility is built, the supercomputer can help keep it operating efficiently too.
A public-private partnership, established in 2009 with funding from the State of Montana, RMSC offers a unique business model that leverages ‘cloud computing’ technology to make HPC more accessible and affordable. RMSC attracted the attention of Northrop Grumman Corporation and in 2009 the two organisations merged their capabilities. HPC infrastructure ‘in the cloud’ with secure, remote on-demand access and a pay-as-you-go pricing schedule has dismantled barriers that once limited supercomputers to big corporations, universities and the federal government.
For example, through 2010, Precision Wind of Boulder, Colorado, ran meteorological algorithms on the RMSC system, dubbed Big Sky, to deliver high fidelity wind forecasts to a commercial renewable energy provider, which used the predictions primarily to prepare for ramping events up to 18 hours in advance. Run even on a large in-house computing infrastructure, it is unlikely these forecasts could have been processed in time.
‘Weather events at this scale are comprised of a nearly infinite number of possible permutations,’ said Robert Kelly, head of business development for Precision Wind. ‘It is impossible to accurately and consistently forecast the most likely outcome without supercomputing capacity, and RMSC’s model allows for flexibility in a potentially very costly environment.’
Having both the flexibility to pay only for processing time used and the option to buy additional processing time as needed may make financial sense for a small, growing services firm like Precision Wind. But looking broadly across the entire energy business, Kurt Paquin, an independent consultant, believes the HPC cloud is the computing future for even the large energy companies that potentially could invest in their own supercomputers. ‘Energy projects take a long time to develop, and computing technology becomes obsolete very quickly,’ said Paquin. ‘There’s a benefit to having a third party with HPC expertise manage the [supercomputing] infrastructure, maintain the technology and keep it up to date.’
Optimising Site Selection
Dr. Randall Alliss, Northrop Grumman’s Atmospheric Sciences manager, explains that the company developed an optimisation algorithm to determine which locations would be best for building communications ground stations based on geographical diversity and cloud cover, with obvious applications in the solar energy industry.
Realising wind farms suffered the same intermittency issues as solar, the team obtained historical wind data and then customised the National Oceanographic and Atmospheric Administration’s (NOAA) Weather Research and Forecasting (WRF) Model. It ran the mesoscale model on the Big Sky system to generate a database of atmospheric conditions for North America so that site selection studies could also be performed.
Although historical wind data are traditionally relied upon for selecting sites, Northrop Grumman saw additional value in predicting how those meteorological and atmospheric conditions will change over time. The team built a regional climate change simulation to determine how evolving climatic conditions will impact the energy generation potential of networks of sites 50 years in the future.
‘These models reduce the risk involved in wind energy development by improving the probability the selected sites have minimal power variances,’ said Kevin MacNeill of Zubalu Inc., a Canadian business development firm working with RMSC.
The power variance score given to potential wind energy projects starts impacting their economic feasibility even before construction begins, explained MacNeill, because investors are always looking to minimise risk. ‘The lowest-risk sites aren’t necessarily those with the highest wind speed; those with the least variance, or intermittency in wind, pose less risk,’ he says. ‘If potential wind farm sites are chosen with power variance scores above 90%, lending institutions will typically offer a 0.5% lower interest rate,’ said MacNeill, adding: ‘On a $100 million project, that translates into a $5 million savings over 20 years.’
The other financial payback from an improved power variance score comes during the operation of the facility. A large variability in wind means that turbines will go through frequent ramping events. Even when predicted through short-term weather forecasts, ramping events add to the operating costs over time. The best remedy is to select a site with low wind variability at the outset. ‘If an operator reduces ramping events by 50%, it can save $10 million over 20 years of operation on a 100 MW project,’ adds MacNeill.
Generating More Power
In the traditional approach to development, several sites are analysed with a combination of historical and predictive wind models. Sites are then ranked in order of potential generation and intermittency. ‘This doesn’t help with the intermittency problem because only one farm is chosen,’ said Alliss. ‘If a developer wants to minimise intermittency, they need multiple geographically diverse farms.’
‘To [select multiple complementary sites], we have to evaluate billions upon billions of different combinations of sites using the supercomputer and a sophisticated search algorithm. We leverage technology from genetic optimisation algorithms to accomplish this,’ said Alliss.
A major operating expense for a wind energy facility is ensuring it has access to power reserves that can accommodate demand during periods of low or high wind. In most cases, the operators must arrange to procure electric power from conventional facilities. Such reserve ‘firming’ contracts are expensive to maintain because they often involve payments even if no power is needed.
Interest in services offering better site optimisation has come from many sectors. Developers, investors and operators of alternative energy facilities find the ability to reduce power intermittency appealing due to the short- and long-term financial benefits it can bring. But state governments and regional energy providers are also looking at the new multi-site selection approach as a way of cutting costs, speeding the permitting process and getting facilities up and running faster.
People are intrigued with ‘the practicality of thinking in a more holistic way of producing power,’ said Alliss.
The bottom line is that any technology which reduces the financing and operating costs of wind energy projects accelerates the return on investment.
Kevin Corbley is a consultant specialising in the geospatial and energy industries.