Güneydoğu Anadolu bölgesi için global güneş ışımasının ve güneşlenme süresinin istatiksel metodlar ile tahmin edilmesi ve karşılaştırılması

Year 2016, Volume: 7 Issue: 1, 73 - 83, 01.06.2016

Hibetullah Kılıç Bilal Gümüş Musa Yılmaz

Abstract

Günümüzde, gelişen teknolojilere paralel olarak enerjiye olan talep de artmaktadır. Artan bu enerji
taleplerinin karşılanması konusunda bilim insanları yeni alternatif enerji kaynakları geliştirmeye
çalışmaktadırlar. Bu çalışmaların büyük bir bölümü ise fosil yakıtlara alternatif olabilecek yenilenebilir enerji
kaynakları üzerinedir. Yenilenebilir enerji kaynaklarından olan rüzgâr ve güneş enerjisi coğrafik koşullara
bağlı olarak, bölgelere göre değişiklik arz etmektedir. Özellikle Türkiye’de güneydoğu Anadolu Bölgesi,
Türkiye ortalaması olan 1400 kWh/m2-yıl’dan daha yüksek, bir güneş enerji potansiyeline sahiptir. Bu
makalede, Güneydoğu Anadolu Bölgesi için, güneş enerjisinin elektrik enerjisine dönüşümünde önemli bir
parametre olan, global güneş ışınımı (GGI) ve güneşlenme sürelerinin (GS) tahmininde, iki farklı istatiksel
yöntem ilk kez kullanılmıştır. Kullanılan yöntemler, üstel ağırlıklı hareketli ortalama (ÜAHO) ve üstel ağırlıklı
hareketli ortalama bazlı gaussian dağılımı(ÜAHOG)’dır. Yapılan çalışmada, global güneş ışınımı ve
güneşlenme süresinin iki farklı istatistiksel yöntemle tahmin edilmesi ve bu yöntemlerin birbirleriyle olan
karşılaştırılması sunulmuştur. Ayrıca yöntemlerin başarı oranı, belirleme katsayısı R2 ve ortalama mutlak
yüzdelik hata(OMYH) kullanılarak test edilmiştir. Yapılan hesaplamalar sonucunda Hem GGI ve GS için
ÜHAO ve ÜAHOG kullanılarak hesaplanan R2 değerleri 1’e yakın olarak bulunurken, diğer taraftan ÜHAO
ve ÜAHOG kullanılarak hesaplanan OMYH değerleri yöntemlerin mükemmel tahmin oranına sahip olduğunu
gösteren 0-10(kWh/m2-gün) aralığında hesaplanmıştır. Elde edilen sonuçlar GGI ve GS tahmini için kullanılan
her iki yöntemin de uygun olduğunu göstermiştir.

Keywords

Global güneş ışınımı, güneşlenme süresi, Üstel Ağırlıklı Hareketli Ortalama (ÜAHO), Gaussian dağılımı

To predict and compare global solar radiation and sunshine duration by using statistical methods for southeastern region of Turkey

Abstract

Dependent of development in technology it causes to
increase in demand of energy. This problem is a
dilemma for science and industry. To overcome this
situation, scientist and researcher try to improve new
solution for energy problem. Solar energy that is a
crucial part of renewable energy is in the interest of
scientists. Especially for regions in the world which
has a high potential of solar energy, it is alternative
solution to fossil fuel. Southeastern Anatolian region
of turkey has rich in view of solar energy. Due to
advantage of this region solar energy may be
replaced to conventional energy types. Because
countries such Turkey that has in progress of
development has to solve their energy problem by
various methods. Although that region of Turkey has
a high potential, it is not enough to benefit from solar
energy.
The aim is this paper to contribute for scientist,
researchers and investigators by eliminating the
disadvantages such as lack of solar parameters
measurement stations due to geographical and
economic problems in that field. The meteorological
measurement station in Turkey are dependent to
Meteorological service of Turkish state. In addition
some station are established in University for
scientific researches. That region of Turkey there is
no enough measurement station. The contribution
proposes in that paper tries to predict solar
parameters by using earlier years’ data.
The important parameters of solar energy are
sunshine duration and global solar radiation. These
parameters are used in conversion of solar energy to
electrical energy. This paper proposes two statistical
method which are used earlier year solar data to
predict next years’ solar parameters such as global
solar radiation and sunshine duration.
First method is exponentially weighted moving
average (EWMA). This method needs minimum two
years’ data for short or long term prediction. Second
method is called as exponentially weighted moving
average (EWMA) based Gaussian distribution
method. It needs also two years’ solar data for
prediction.
Both mentioned method are applied to five cites of
southeaster Anatolia region of Turkey that are
Gaziantep, Şanlıurfa, Diyarbakır, Batman and
Mardin. The two years’ (1998-2000) data are used
for test class years’ data in Gaziantep, Şanlıurfa,
Diyarbakır, Batman but in Mardin two years’ data
used for test class years are between 2012-2013. The
predicted years for Cities of Gaziantep, Şanlıurfa are
10 years, for Diyarbakır are 8 years, for Batman are
6 years which are long term prediction. However for
Mardin cities predicted years are 2 years and it is
called as short term prediction. It is a obligation to
selected this years for prediction because only that
years solar data can be obtained from metrological
service of Turkish state.
For both method two statistical tools which area
mean average percentage error (MAPE) and
determination coefficient (R2
). The computed MAPE
for global solar radiation and sunshine duration in
EWMA method is between 0-10(kWh/m2
) that is
excellent prediction accuracy and the computed R2
for global solar radiation and sunshine duration is
close to 1 which is indicated measured and predicted
data are similar. The computed values of MAPE for
EWMA based Guassian distribution method is also in
acceptable range that is between 0-10(kWh/m2) for
global solar radiation. On the other hands they are
also indicated excellent prediction accuracy for
sunshine duration excluding Diyarbakır and Batman
cities it is indicated good prediction accuracy which
is in the interval of 10-20(kWh/m2). The computed
values of R2 for both global solar radition and
sunshine duration in EWMA based Gaussian
distribution method is close to 1 and it means the
predicted data are fitted with measured data.
As a result EWMA method and EWMA based
Gaussian distribution method have high prediction
accuracy. However the result of EWMA methods has
a higher prediction accuracy than values of EWMA
based Gaussian distribution method.

Keywords

Global solar radiation, sunshine duration, EWMA, Gaussian distribution