Soft Computing Techniques applied to the wind forecast on wind farms

Description of the technology

The research group, Modern Optimization Heuristics and Communication Network Design (GHEODE), of the University of Alcala, offers the use of modern soft computing techniques, deeply tested, to generate automatic systems that will allow performing wind in wind farm. With this technique energy operators could minimize the impact of the penalties for the lack or excess of produced energy and hence maximize the benefit.

Specifications

Soft Computing is one of the key research lines in the artificial intelligence world. It focuses, among other things, on the design of smart systems that can obtain and represent information given by large information data series. These series could contain some contradictions, incoherences, could be incompletes or even with low precision. Using soft computing techniques for this task allows obtaining much better and robust solutions. At the same time they allow to reduce the cost associated to the time (of obtaining the solutions with classical techniques). In our research group we have focused our work in three soft computing techniques: Neural Networks, Support Vector Machines and Evolutive Techniques. The first two could be considered the most important research lines in Soft Computing in the last years. The first is based in the simulation of the behaviour of the human brain for event learning and their subsequent reproduction. The second one is based on complex statistics models for the search of new models that give better performance to future external inputs. Both techniques allow the development of regression models based on data obtained from sampling the real world process, in such a way that allow to replay or forecast approximately the real world process. Evolutive techniques allow solving complex optimization problems in a fast and efficient fashion. Among other things, they allow to do changes over the above regression models to enhance their performance in such a way that it will be almost impossible to do with classical techniques One of the clearest applications where we have applied these techniques is the wind forecast in wind farms. To do this, given some data series of variables of the wind farm environment, it is possible to built regression models for make forecast of the wind strength in each one of the wind turbines of the specific wind farm. These variables could be time invariant, as the orography around the wind farm, past time variables, as information about the wind in the turbines in past years, or meteorological variables (temperature, wind in a specific point outside the farm…), obtained using meteorological global models. Obtaining the forecast of the wind in each turbine allows evaluating the total energy generated in the wind farm, not only in normal working conditions but also when some of the turbines are broken or in a repair process. Furthermore, it is possible to hybridize these regression models with other meteorological models or even with the network grouping techniques for performance enhancement.

Main advantages of its use

• Any enhancement in the use of renewable energy is a good thing from the environmental point of view, because it allows reducing the use of other polluting technologies. Moreover it is very profitable for the companies that will acquire them, because of the reduction of the economic penalties of the governments for the error in the energy forecast in the wind farms.

Applications

• Alternative energy production and wind energy.