# Introduction

he troposphere is the lower layer of the Earth's atmosphere. Most of the weather phenomena, systems, convection, turbulence and clouds occur in this layer, although some may extend into the lower portion of the stratosphere. The troposphere contains almost all the atmospheric water vapour. It contains about 70 to 80 percent of the total mass of the earth's atmosphere and 99 per cent of the water vapour.

Temperature and water vapour content in the troposphere decrease rapidly with altitude and thus most of the water vapour in the troposphere is concentrated in the lower, warmer zone. Water vapor concentrations vary with latitude. The condition of the atmosphere as dictated by the sun is very dynamic both in space and time scales. The resulting solar interactions on the atmosphere leads to changes in weather as well as the so called climate change. 


# Year

The objective of this study is to model the relevant data provided by the Nigerian Meteorological Agency, Federal Ministry of Aviation, Oshodi, Lagos, Nigeria as shown in Table 1 and the global solar radiation data collected from Renewable Energy for Rural Industrialization and Development in Nigeria using the Angstrom-Prescott, statistical technique (multiple regression model) and the artificial neural networks (ANN), and then comparing the results of these three models along with the measured solar radiation data. The table below (Table1) shows the atmospheric parameters in preparation of the prediction.
Ibeh G.F ? , Agbo G.A ? , Oboma D.N ? , Ekpe J.E ? , & Odoh S II.

# Methods and procedures Multiple regressions

Regression method is one of the most widely used statistical techniques Mendenhall and Beaver (1994) and Sharda, and Patil (1990). Multiple regression analysis is a multivariate statistical technique used to examine the relationship between a single dependent variable and a set of independent variables. The objective of the multiple regression analysis is to use independent variables whose values are known to predict the single dependent variable (Olaniyi, 2011). The effect of independent variables on the response is expressed mathematically by the regression or response function f:
y = f (x 1 , x 2 , ?, x n ; ?? 1 , ?? 2 , ?, ?? n ) (1)
where y is the dependent variable.

? 1 , ? 2 , ?, ? n in the regression parameters which will be determined

The regression model for the observed response variable is written as
?? = ?? + ?? = ð??"ð??" ??? 1, ?? 2, ? , ?? ?? ; ?? 1, ?? 2 , ? ?? ?? ? + ?? (2)
Where ?? is the error in observed value z (Olaniyi, 2011).

To find unknown regression parameters {?? 1, ?? 2 , ? ?? ?? }, the method of least squares ( Beenstock, 1992) can be applied:
E (?? 1, ?? 2 , ? ?? ?? ) = ? (?? ?? ? ?? ?? ) 2 = ? (?? ?? ? ð??"ð??" (?? 1 ?? 2 , ? , ?? ?? ; ?? 1, ?? 2 , ? ?? ?? ) ) 2(3)
Where E (?? 1, ?? 2 , ? ?? ?? ) is the error function or sum of squares of the deviations.

To estimate ?? 1, ?? 2 , ? ?? we minimize, E, by solving the system of equations:
???? ???? 1 = o, ?? = 1,2, ? , ??(4)

# a) Model Specification and Analysis

The regression model to consider in this study takes the maximum temperature (Tm) and sunshine hour (?? ? ) as the explanatory variables and measured values (?? ?? ) as dependent variable. This is used to obtain a reliable parameter estimates in the regression. The model to be used can be specified as
?? ?? = f (?? ?? , ?? ?) (5)
More precisely;
?? ?? = ?? 0 + ?? 1 ?? ?? + ?? 2 ?? ? (6) ? 0 , ? 1 , ? 2 > 0 III.

# Angstrom-prescott model

There are several types of empirical formulae for predicting the monthly mean daily global solar radiation as a function of readily measured climatic data (Iqbal, 1977;and Klien, 1977). Among the existing correlations, the one used in this paper is the Angstrom-Prescott regression equation, which relates the monthly mean daily global solar radiation to the number of hours of bright sunshine as follows The monthly mean daily extraterrestrial radiation ?? ? ?? and monthly mean day length ?? ? was derived from the formulae: 
?? ? ?? = 24 ?? ?? ???? ?? ?? [?? ?? sin ?? sin ?? + cos ?? cos ?? cos ?? ?? ] ?? ??? = 2 25 COS-1 (-tan ? tan ?)(9)? = 23.45 sin [ ??+284 365 ](10)?? ?? = COS -1 (-tan ? tan ?)(11)
IV.


# Artificial neural network model

Artificial neural network models are based on the neural structure of the brain . The brain learns from experience and so do artificial neural networks. Previous research has shown that artificial neural networks are suitable for pattern recognition, prediction and pattern classification tasks due to their nonlinear nonparametric adaptive-learning properties. As a useful analytical tool, ANN is widely applied in analyzing the data stored in database or data warehouse nowadays (Massie, 2001). One critical step in neural network application is network training. Generally, meteorological data is selected and refined to form training data sets. Artificial Neural Network is widely used in various branches of engineering and science and their unique proper of being able to approximate complex and nonlinear equations makes it a useful tool in quantitative analysis (Olaniyi, 2011).

The true power and advantage of neural networks lies in their ability to represent both linear and Traditional linear models are simply inadequate when it comes to modeling data that contains non-linear characteristics. In this paper, one model of neural network is selected among the main network architectures used. The basis of the model is neural structure as shown in Fig. 1. These neurons act like parallel processing units.


# a) Multi-Layer Perception

The most popular network architecture in use today is perhaps the Multilayer Perceptron network and it uses supervised network (Hair and Tatham, 1998). This type of neural network is known as a supervised network because it requires a desired output in order to learn. The goal of this type of network is to create a model that correctly maps the input to the output using historical data so that the model can then be used to produce the output when the desired output is unknown. Figure 1 show the block diagram of a double hidden layer multiplayer perceptron (MLP). The inputs are fed into the input layer and get multiplied by interconnection weights as they are passed from the input layer to the hidden layer. Within the hidden layer, they get summed then processed by a nonlinear function (usually the hyperbolic tangent).

If more than a single hidden layer exists then, as the processed data leaves the first hidden layer, again it gets multiplied by interconnection weights, then summed and processed by the second hidden layer and so on. Finally the data is multiplied by interconnection weights then processed one last time within the output layer to produce the neural network output.

An artificial neuron is a unit that performs a simple mathematical operation on its inputs and imitates the function of biological neurons and their unique process of learning . The weighted sum of the inputs are calculate using the following equation, (7) Where v k is the weight sum from the kth hidden node, ?? ???? is the weight on connection from the jth to the kth node; x j is an input data from input node; m is the total number of input (N=17); and b k denotes a bias on the kth hidden node.

The mathematical structure of the normal method is as shown in fig. 2 Output layer Hidden layer Input layer V.
? ? ? ? ? ? ( D D D D ) D

# Results and discussions

Table 2 is obtained from the SPSS output for the analysis of the multiple linear regressions relating the measured values of global solar radiation as a function of the maximum temperature (Tm) and the sunshine hour (n). The standard error for each of the variables is indicated in the table 3. From figure 1, the highest value of solar radiation is 18.16 and 18.11MJ/M 2 /Day respectively. These results suggest that there is peak dry season in Warri during December and April when the solar radiation is high. Again, the low values of solar radiation occur from May to August, which indicates peak rainfall Warri when the sky is cloudy and solar radiation is low. 


# Conclusion

In this paper, three techniques for modeling and forecasting the solar radiation of Warri Nigeria: Neural Network, Angstrom-Prescott and Statistical Technique. The forecasting ability of these models is accessed on the basis of MBE, RMSE and MPE. We have discovered the fact that Neural Networks output perform better in forecasting from table 2 and 3 compare to other two models. Thus, ANN should be used for prediction of global solar radiation of the location and other location that has similar condition. 


# ERROR ANN REGRESSION


# ANGSTROM-PRESCOTT
![?? ? ?? = measured monthly mean daily global solar radiation on a horizontal sunshine. ?? ? =monthly mean daily bright sunshine hour, ?? ? = Maximum possible monthly mean daily sunshine, a and b = regression constant.](image-2.png "")
![?? ???? = solar constant (4.921 MJM -2 day -1 ) N = characteristic day number ? = latitude angle ? = angle of declination a = -0.110 + 0.235 cos? + 0.323 (n/N) (12) b =1.449 -0.553 Cos ? -0.694 (n/N)](image-3.png "")



1T( D D D D ) DMonth?? ? HourT m ??? ? ?? (MJm 2 day -1 )JAN4.7233.0012.52FEB4.8033.6814.26MAR4.6133.4515.64.APR4.9232.8618.11MAY4.8931.9312.84JUN3.8630.5313.99JUL2.2728.7714.67AUG2.3128.8913.85SEPT2.5729.9915.40OCT4.1531.2816.55NOV5.2332.7415.81DEC5.6632.6618.16
2Variable CoefficientStd.t-statistic Prob.Std.ErrorTm-0.3620.656-0.4610.785?? ?0.9660.4280.8300.617Constant22.5910.3011.0980.000
3
		
		
			

				


* 
	
		
		
			MBeenstock
		
		
			EGoldin
		
		
			DNabot
		
	
	
		Energy Economics
		
			21
			
			1990
		
	




* 
	
		Multivariate Data Analysis
		
			AHair
		
		
			BTatham
		
		
			1998
			Pearson Education press
		
	




* 
	
		Neural Network fundamentals for scientist and engineers
		
			DDMassie
		
	
	
		Proceedings of the international conference of efficiency, Costs, Optimization, Simulation and Environmental Aspects of Energy systems and Processes
				the international conference of efficiency, Costs, Optimization, Simulation and Environmental Aspects of Energy systems and ProcessesIstanbul Turkey
		
			2001
			
		
	




* 
	
		Estimation of monthly global solar radiation for Warri-Nigeria, using Angstrom and ANN models
		
			GFIbeh
		
		
			GAAgbo
		
	
	
		Advances in Science research
		
			3
			1
			
			2012
		
	




* 
	
		Comparison of empirical and artificial neural network models for the correlation of monthly average global solar radiation with sunshine hours in Minna
		
			GFIbeh
		
		
			Agbo
		
		
			Ga
		
		
			SRabia
		
		
			Chikwenze
		
		
			Ar
		
	
	
		Nigeria. International Journal of physical Sciences
		
			7
			8
			
			2012
		
	




* 
	
		An Introduction to Solar Radiation
		
			MIgbal
		
		
			1983
			Academy Press
			New York, NY
		
	




* 
	
		Calculation of Monthly Average Isolation on Tilted Surfaces
		
			SAKlein
		
	
	
		Solar Energy
		
			19
			
			1977
		
	




* 
	
		Introduction to Probability and Statistics. Ninth Edition, International
		
			BeaverMendenhall
		
		
			1994
			Thomson Publishing
		
	




* 
	
		
		
			NeuralworksNeuralware
		
		
			Predict
		
		
		
	




* 
	
		Comparison of Regression Model and Artificial Neural Network Model for the prediction of Electrical Power generated in Nigeria
		
			SMOlaniyi
		
		
			OAAnthony
		
		
			OMAdeola
		
		
			M ILawrence
		
		
			OAChukwuemeka
		
	
	
		Advances in Applied Science Research
		
			2
			5
			
			2011
		
	




* 
	
		Neural Networks as forecasting experts: an empirical test
		
			RSharda
		
		
	
	
		Proceedings of the International Joint Conference on Neural Networks
				the International Joint Conference on Neural NetworksWashington DC, USA
		
			SPSS Inc
			1990. 2008
			I
			
		
	
	SPSS17 software