Geology Reference
In-Depth Information
Fig. 5.30 Solar radiation as observed and estimated using the ANFIS model for the training data
set. a scatter plot, b line diagram
over 24 h), and daily rainfall (summed over 24 h), and the output layer contained a
single daily solar radiation. The subtractive fuzzy clustering was used to establish the
rule-based relationship between these six input data series and output data variable.
The Subtractive clustering automatically identify the natural clusters in the input-
output data pool. In this ANFIS model, there were 54 parameters to determine in the
layer 2 because of 6input variables, 3 rules. The three rules generate 3
6
nodes in
subsequent layer. The study set the number of membership functions for each input
of ANFIS as three with Gaussian (or bell-shaped) and linear membership functions
at the inputs and outputs respectively. The accuracy of the modelling by ANFIS was
assessed through variety of statistical parameters like namely CORR, Slope, RMSE,
MAPE, MBE, ef
ciency and Variance of the distribution of differences about MBE,
(S
2
). Predicted daily solar radiation of the Brue catchment are compared with the
corresponding observed values during training phase (both scatter plot and line
diagram) in Fig.
5.30
a, b; the corresponding results during the validation phase are
shown in the Fig.
5.31
a, b.
The study has used a state of the art support vector machines (SVMs) for daily
solar radiation modelling. SVMs are considered as an integral part of studies on
pattern recognition and are essentially a major sub discipline of machine learning.
The study has used LIBSVM with capabilities modelling of
ʽ
-SV regression and
Fig. 5.31 Solar radiation as observed and estimated using the ANFIS model for the validation
data set. a scatter plot, b line diagram
