Variability is a well-known characteristic of windpower, and system operators know they must plan for changes in wind generation over the c
The Wind Watcher
wind behavior have changed and therefore the model not only learns the new patterns of behavior but also forgets knowledge previously learned, whose consideration may now be more harmful than good.
We are also applying new concepts related to information entropy in the training process of our intelligent systems, instead of using a classic approach based on the variance of the prediction errors.
And we are using the full power of information gathered from individual wind turbines instead of trying to predict the behavior of a wind farm as a whole, having built up a process of naturally taking into account unit outages or unit additions of any distinct rated power.
Fortnightly: What is the role of Horizon Wind Energy in the project?
Miranda: Horizon Wind Energy is an extremely important partner that will help in validating our work. We already have the experience of working with the parent company of Horizon Wind Energy, EdP. In this country, we have worked in a consortium, EPREV, of companies and research institutions to build up a dispatch center for wind energy, to serve the interests of wind energy producers. This has certainly created a link of confidence and also of responsibility that will allow ANL and also INESC Porto to submit and discuss the work with business actors. Horizon may also serve as a source of data and as test bed for the models to be developed. ANL has made sure there would be one such partner in the project and this is a good thing. I would like also to make a reference to another important partner, with a distinct yet no less important point of view, which is the Midwest ISO.
Fortnightly: Will the resulting wind forecasting system that is developed be proprietary?
Miranda: The policy will be defined by Argonne National Laboratory and I am not entitled to issue an opinion on this matter. But it is my understanding, and we are working with this motivation, that we are building not only a demonstrator but also a robust tool that may be used as an independent benchmarking platform. In Europe there is no such independent platform. This will provide for wind generation companies and also for system operators, a means to independently assess the quality of the prediction models and processes they have, or will, adopt. Hopefully, this will also be seen as an incentive for actors in a wind power prediction industry to improve the quality of the models they offer.