The assessment of the suitability of a wind system
depends largely on the prediction of the wind potential. Indeed,
the variability and uncertainty inherent in renewable energy
sources can have a significant impact on accurate and reliable
prediction of the power produced. Wind sources are needed at
different time stages and at different altitudes. Thus, putting in
place tools for predicting these wind resources is essential for
their effective integration in the frame of electricity generation.
In this context, the paper of this study is to propose a short-term
wind energy prediction method through the formation of
historical wind velocity data based on neural networks. This
assessment involves modelling wind speed using ANN through
the feed-forwad network. So, ANN are at the basis of adaptive
identification methods and intelligent command laws. In this
sense, first, the process of forecasting wind energy involves the
creation of a raw data base, which is then filtered by probabilistic
neural network. More concretely, the contribution of the work
can be given in the form of technical results. These results start
with a proposal of the theoretical models, then it is given the
approach method that is used, then it is proposed the design of
the system and the whole is closed by a performance evaluation.
As far as performance evaluation is concerned, it is presented in
the form of the results of analysed simulations of the forecast
model. In practical terms, it should be noted that the proposed
model also provides a high degree of accuracy for the measured
data. In the end, normalized average absolute errors were
recorded between 4.7% and 4.9%. As, it was found a regression
factor R (measures the correlation between output-Target)
between 91% and 96% for the site of the northern Mauritanian
coast. This is largely acceptable for similar calculations.
Neuron Network Prediction Feed-Forwad Wind Speed Network on Mauritania's North Coast: Ballawack Case
