Sunday, November 2, 2014

WindWX - A Wind Forecast in the Realm of Supply Chain Managment

When I hear somebody talk about a good "forecast", I instantly think of a warm and sunny day, eagerly anticipating the taste of my favorite Ben & Jerry's (cookie dough!) on my tongue. 
For Xcel Energy's 3,000 wind turbines, this expectation might be translated into their blades' thirst for winds and gusts. And Xcel more and more excels at predicting when their needs will be met: the energy company managed to save its customers fuel of up to $40 million during the past four years.

As we know from Professor Wheelwright (Brigham Young University), forecasting makes sense if and only if it is possible to deduce a particular action from doing so AND if it proves reliable, which implies a sufficient amount of supporting data. 

The specific action for a big wind energy company to take as a result of useful forecasting is to shut down those wind parks that are, relatively to other wind parks, less efficient under the respective (forecasted) circumstances - if these plants are able to produce enough energy to meet the overall demand. By doing so, the company allocates its resources (labor, energy for the operation of the wind turbines, etc.) more efficiently and can better balance out fluctuations that are due to the volatility of wind. 

Rather problematic is the fulfillment of the second requirement. For a long time, it seemed to be a sheer impossible task to develop algorithms capable of forecasting wind. In order to overcome this obstacle, Xcel cooperated with Global Weather Corp. to model WindWX, a wind-production forecasting system. WindWX works as follows: by processing real-time, turbine-level operating data, it computes which of the many complicated algorithms it has to apply to make a reasonable 15-minute-period forecast about the amount of wind power the plants will generate. 

Xcel claims that the forecasting model could be used by any wind energy company; its unique selling point, however, is its large number of wind parks, making Xcel the No. 1 wind energy provider in the United States. The vast amount of its wind turbines allows it to dispatch its systems according to the forecast, while smaller companies would not able to act accordingly (thereby failing the precondition for using forecasting at all). By refining its algorithms, Xcel will be able to expand its lead even further, which puts the other companies at an increased disadvantage.

To me, this sounds nearly too good to be true. How often did it happen that I was looking forward to enjoying my ice cream, only to find myself - shivering - opening my umbrella in the end... And now there is this company that invests a few millions of $ in research and reduces its forecasting error rate by more than 30 percent, somehow solving one of nature's mysteries. I am just wondering, though, what the initial error rate was. Although savings of $40 million for the end consumer sound pretty impressive at first, this might be just the tip of the iceberg relative to the actual potential that could be realized. Also, I assume that algorithms can expire - due to global warming, a decrease of wind as a consequence of higher temperatures might affect the long-term reliability of the forecasting system. And there is another question that comes to my mind: how can the narrow time frame of 15 minutes of forecasting lead to the decision to power down particular wind turbines? If wind turbines are completely shut down, it takes them somewhere between 10 and 30 minutes to start up again. This results in a loss of flexibility if the forecast where to change for the subsequent 15-minute period... 


About Xcel and its forecasting model:
About the relationship between climate change and wind:
About the start up time of a wind turbine:
Clipart: Photo:

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.