Roberto Mejia ECET 5214-17-2012 The purpose of this paper is a time series prediction of the wind speed with artificial neural networks are presented. The paper written is called Time Series Prediction of Wind Speed written by Soteris

madbrainedmudlickΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

64 εμφανίσεις

Roberto Mejia

ECET 521

4
-
17
-
2012



The purpose of this paper is a time series prediction of the wind speed with artificial neural networks



are presented. The paper written is called Time Series Prediction of Wind Speed written by Soteris



Kalogirou and Silas Michaelides. Wind speed prediction is very important for the long term estimation


of the performance of wind turbines. The increase use of wind energy and the depletion of the fossil



fuel reserves combine with the increase of the env
ironmental pollution have encourage the search for


clean and pollution
-
free sources of energy.



Data was collected from a region located at the southern part of Cyprus. The observed data of wind


speed in the study covered ten consecutive years (1991
to 2000). A training data set having wind speed


records for each hour of a month for the years from 1991 to 1999 and verification data set for all the


hours of each month for the year 2000. The training data set has been used for the training and


testi
ng of the artificial neural network, while the verification data set has been used to validate of the


network. The network was trained to predict the mean monthly hourly wind speed of the year (1994) by


using the values of wind speed, for the same month

and same hour, for the previous three years (1991
-


1993), then the data for the years 1992
-
1994 were used as input and those for the year 1995 as output


and so on. The procedure is used consecutively for the preparation of the required data bases. The


data that was collected for the year 1997
-
1999 (input) and 2000(output) were used for the validation of


the network.



Different network structures, sizes and learning parameters have been tried. One unit is composed


of five slab, three of which are

hidden. The five neuron input layer comprise the input data. These


are the month, hour and mean monthly hourly wind speed for three years. The output of the network is a


single neuron representing the mean monthly hourly wind speed for the next year.



This paper had two different network structures and of the two I would choose the first network



discussed in the summary. The first network is base on real real numbers and has years of data


were I feel that the second network discussed has too man
y variables that can alter the results.



This paper is a time series prediction of wind speed with artificial neural network is presented. Wind



speed records from the area of Kourris dam, located south of Cypus was used. Data that was used



were wind speed records of the month hourly mean wind speed of the years (1991
-
1999). I


agreed with the first network chosen that it has enough data to conclude a accurate result while


the second network didn’t have enough data to support its theory.