^{1}

^{*}

^{1}

The importance of accurate knowledge about available global solar radiation in the design and development of various solar energy systems cannot be overemphasized. Most of the available models for predicting global solar radiation involve a plethora of input factors, some of which require special skills and equipment to measure. Such multi-factor models are complex and computationally demanding. To remove some burdens associated with such models, the use of simplified prototypes with reduced input factors ha s been proposed. It has been shown that a model with fewer input factors, that can be determined in a definite manner or whose attributes are directly observable, is often a better alternative. Therefore, the main object of this paper is to have models with a few variables that can easily be measured, developed for predicting global solar radiation. Two input factors , geographical location and season of the year , were considered. Using a 22-year interannual average daily insolation data from the database of the National Aeronautics and Space Administration (NASA) blended with the art of interpolation, empirical models were fashioned with the data for the five subregions of Africa. The results of the models’ analysis indicate that the latitude component is the dominant locational factor. Furthermore, the new models exhibit optimal performance in comparison with existing models and constitute reliable predictive tools that are suitable for estimating global solar radiation for any practical application.

Energy is invisible, nonetheless a driving force in the entire universe. It is a fundamental input to any human activity [

In the 19^{th} century, the use of fossil fuels enhanced rapid industrialization and modern civilization. However, excessive exploitation of fossil fuels directly and indirectly assists in global warming and other encumbrances which drive our planet towards a dark future. Today, technological advancements have enabled the development of alternative and renewable energy sources, one of which is solar energy.

Solar energy is free, it does not create pollution, and has helped man become less dependent on other more costly and damaging forms of energy [

Generally, people understand real-world phenomena better when they are represented symbolically. As such, modelling should be used not only to illustrate and deepen understanding, but also to predict real-world phenomena [

Most of the available models for predicting global solar radiation use some or even all of the following factors as input: sunshine data, solar declination, latitude, longitude, extraterrestrial radiation, relative humidity, soil temperature, temperature of the air, cloudiness, evaporation and precipitation among others. These factors require special skills and equipment to measure. In addition, models with multiple inputs have been shown to have insignificantly improved in accuracy compared with variants of the models in which fewer inputs are considered. Indeed, such variants are simpler to handle, not only in terms of computational convenience but also for ease of analysis, see Latunde et al. [

Therefore, the main objective of this paper is to develop models, with few input factors that can be used to predict global solar radiation. The new models introduced in this study require little computational effort, and the only information needed to use them is the knowledge of the longitude and latitude, in addition to the season at the location of interest.

Solar energy refers to sources of energy that can be directly attributed to the light of the sun or the heat that sunlight generates. It is also a fact that solar radiation is electromagnetic in nature. Global solar radiation, on the other hand, is the sum of the direct, diffuse and reflected solar radiations [^{2}). Direct estimates of solar energy may be expressed as watts per square meter (W/m^{2}) while those for heating systems are usually measured in British thermal units per square foot (Btu/ft^{2}).

Several models have been proposed for estimating daily or monthly global solar radiation using different techniques such as geostationary satellite images, neural networks, time series methods, physical radiative transfer models, and stochastic weather methods. These are generally based on different types of data including meteorological and geographical data [

Meteorological data-based models depend on the most common meteorological elements including cloud cover, sunshine duration, temperature, and relative humidity, making them the most widely used models, especially the sunshine-based and temperature-based models [

Many solar energy researchers have adopted Angstrom-Prescott-Page model as a baseline to further develop empirical models for predicting global solar radiation. One of these researchers is Falayi [

In other locations in Africa, Soufi et al. [

It is pertinent to note that majority of the variables involved in most of the studies listed above require special skills and equipment to measure, whereas there are other variables which can be simply determined in a definite manner, and whose attributes are directly observable. Two of such factors and which are proposed for use in this study, are geographical location and seasonal variation.

Although the sun’s energy output is fairly constant, the total solar radiation falling on the surface of the earth varies and depends on a number of factors. According to Mittasova [

For this study, the geographical factors, latitude and longitude, are considered in modelling global solar radiation simply because every location on earth receives sunlight at least part of the year and the amount of solar radiation that reaches any spot on the earth’s surface basically varies according to the geographical location and season. Both latitude and longitude are the means by which the position of any place on the earth’s surface are determined and their attributes directly observed.

In this section, basic issues relating to the base and methodology of the study are considered. The domain for this application encompasses all the nations of the world, but the African continent has been selected for illustrating the methodology adopted.

Geographically, Africa lies between latitudes 37˚N and 35˚S, and between longitudes 52˚E and 17˚W. The continent comprises of 54 independent countries. In assessing climate features of geographical locations and their seasons in this study, the African continent was subdivided into the five approved subregions, namely, North Africa, West Africa, Central Africa, East Africa and Southern Africa. These zones are shown in

North Africa is characterized by the Mediterranean climate at the coast and a large desert area in the south where temperatures are hottest. The seasons observed in the subregion are Spring (March-May), Summer (June-September), Autumn (October-November) and Winter (December-February) [

The West African subregion is influenced by the inter-tropical divergence. The rainy season in coastal areas is generally observed from the end of April to July with a second and shorter rainy season in September and October. Further inland, only one rainy season is observed from June to September.

Central Africa is one of the wettest parts of the continent. Three peak rainfall periods are experienced in the region, namely, March-May, July-September and October-December.

The climate in East Africa is typically equatorial with high temperatures year round and little seasonal variation especially close to the Equator. There are two short rainy seasons: one around April-June, and the other from October-December.

The Southern Africa subregion is mainly characterized by a wet and hot season from October to March (Summer), and a cool and dry season from April to September (Winter).

There are largely two broad categories of techniques, with varying structures, available for developing mathematical models; these are the theoretical and empirical modelling [

The empirical technique, on the other hand, makes use of previous measurements that reveal relationships between the variables. Such relationships are to be utilized to make some predictions. Empirical modeling was, therefore, adopted for use in this study since the theoretical framework for forming some relationship between the factors is not available.

To generate the empirical data needed for this study, latitudes to the north of the Equator are regarded as positive while those to the south of the Equator are taken as negative. Similarly, longitudes to the east of the Greenwich Meridian are assigned positive values while negative values are attached to longitudes to the west of the Greenwich Meridian.

Interannual monthly average daily insolation for the period between 1996 and 2017 from the database of the National Aeronautics and Space Administration (NASA) were obtained and processed to average global solar radiation, being guided by the information from the World Meteorological Organization on the climatic conditions and seasons, over the entire five subregions in Africa.

Locations were carefully selected within each of the subregions and the average global solar radiation were collated via the Power Data Access Viewer [

The method of interpolation was adopted in developing global solar radiation models for any geographical location in the subregions.

In handling discrete data, with the intent of generating a predictive tool, it is customary to apply the concept of interpolation. Polynomials are commonly used for interpolation because they are easy to evaluate, differentiate and integrate [

The Wolfram mathematical software is a typical tool that is designed for constructing interpolants that are capable of evaluating a function and its derivatives at specific interpolation points [

North Africa:

R N 1 = 12.0388 + 0.12609 x − 0.0182203 x 2 + 0.000271859 x 3 − 0.28832 y + 0.0141218 x y − 0.000153931 x 2 y − 0.000148864 y 2 R N 2 = 2.5493 + 0.509602 x − 0.0156812 x 2 + 0.000128806 x 3 + 0.0604844 y − 0.0059339 x y + 0.00016503 x 2 y − 0.000774431 y 2

R N 3 = 11.1332 − 0.0107351 x − 0.0136715 x 2 + 0.00021861 x 3 − 0.232581 y + 0.0127298 x y − 0.000167089 x 2 y − 0.000233522 y 2 R N 4 = 10.573 − 0.0702876 x − 0.00987517 x 2 + 0.000162251 x 3 − 0.157611 y + 0.00880711 x y − 0.000120762 x 2 y − 0.000516666 y 2 (1)

where R_{N}_{1},R_{N}_{2},R_{N}_{3},R_{N}_{4} denote the seasonal global radiation for the four seasons in North Africa, namely, Spring, Summer, Autumn, and Winter. The domain of each of the four models is

D N = { ( x , y ) : 5 ≤ x ≤ 40 , − 10 ≤ y ≤ 40 } .

West Africa:

R W 1 = 3.96376 + 0.0228769 x + 0.0158509 x 2 − 0.000421256 x 3 + 0.00587765 y − 0.00199591 x y + 0.00015084 x 2 y − 0.00134874 y 2 R W 2 = 4.77514 + 0.0707062 x + 0.0057673 x 2 − 0.000265387 x 3 + 0.00052142 y + 0.00487346 x y − 0.00021861 x 2 y − 0.000433544 y 2 (2)

where R_{W}_{1} and R_{W}_{2} denote the seasonal global radiation, for the two seasons in West Africa, in the domain D W = { ( x , y ) : 0 ≤ x ≤ 30 , − 20 ≤ y ≤ 20 } .

Central Africa:

R C 1 = 2.94796 − 0.0182047 x + 0.0158222 x 2 − 0.000301331 x 3 + 0.144672 y + 0.00399308 x y − 0.000377977 x 2 y − 0.00260641 y 2 − 0.0000226986 x y 2

R C 2 = 3.84798 − 0.0645648 x + 0.0224157 x 2 − 0.000258392 x 3 + 0.0771062 y + 0.0131217 x y − 0.00086798 x 2 y − 0.000516168 y 2 − 0.000325 x y 2 (3)

where R_{C}_{1} and R_{C}_{2} denote the seasonal global radiation for the two seasons in the domain

D C = { ( x , y ) : − 15 ≤ x ≤ 25 , 10 ≤ y ≤ 32 } .

East Africa:

R E 1 = − 2.88039 + 0.450858 x + 0.00583204 x 2 + 0.000164107 x 3 + 0.622334 y − 0.0236714 x y − 0.0000878816 x 2 y − 0.0155337 y 2 + 0.000275662 x y 2 + 0.00013103 y 3 R E 2 = 1.27347 + 0.350915 x + 0.00404128 x 2 + 0.0000642684 x 3 + 0.313112 y − 0.0183783 x y − 0.0000775406 x 2 y − 0.0074416 y 2 + 0.0002264 x y 2 + 0.0000622583 y 3 (4)

where R_{E}_{1} and R_{E}_{2} denote the seasonal global radiation in the valid domain

D E = { ( x , y ) : − 25 ≤ x ≤ 18 , 30 ≤ y ≤ 55 } .

Southern Africa:

R S 1 = 3.40494 − 0.0490626 x − 0.00285428 x 2 + 0.0000219348 x 3 + 0.196849 y − 0.00132395 x y − 0.00491978 y 2 R S 2 = 25.7696 + 2.42097 x + 0.0954067 x 2 + 0.00122414 x 3 + 0.0208147 y − 0.00461178 x y − 0.00414814 y 2 (5)

where R_{S}_{1} and R_{S}_{2} denote the seasonal global radiation for the two seasons in Southern Africa, together with their domain D W = { ( x , y ) : − 40 ≤ x ≤ − 18 , 10 ≤ y ≤ 35 } .

The generalized form of the models given by Equations (1)-(5) is

R G k = a + b x + c x 2 + d x 3 + e y + f x y + g x 2 y + h y 2 + i x y 2 + j y 3 (6)

where a , b , c , d , e , f , g , h , i , j ∈ R , − 2.89 < a < 25.77 , − 0.08 < b < 2.43 , − 0.02 < c < 0.1 , − 0.0005 < d < 0.002 , − 0.29 < e < 0.7 , − 0.03 < f < 0.02 , − 0.0009 < g < 0.0002 , h ≤ 0 , − 0.0003 < i < 0.004 , j ≥ 0 ; G = N,W,C,E or S and k = 1,2,3 or 4.

Contributions of location components x and y in determining the average seasonal global horizontal insolation in subregions of Africa as shown in

Step 1: Divide each model R_{Gk} through by its least coefficient.

Step 2: Using the domain of definition D_{G}, neutralize x and subsequently y in R_{Gk}.

Step 3: Determine the dominant location component and its percentage contribution.

Mathematical models (1)-(5) were also used to predict values of global solar radiation for the latitude and longitude coordinates of selected places in each subregion. The results for the predicted and actual values are as shown in Figures 8-12.

The coefficients of correlation [

The location-based models by Gopiathan [

Region | Domain D_{N} | Model | % contribution | Dominant component | |
---|---|---|---|---|---|

x | y | ||||

North Africa | 5 ≤ x ≤ 40 | R_{N}1 | 87.23 | 12.77 | x |

−10 ≤ y ≤ 40 | R_{N}2 | 63.60 | 36.4 | x | |

R_{N}3 | 60.99 | 39.01 | x | ||

R_{N}4 | 39.79 | 60.21 | y | ||

West Africa | 0 ≤ x ≤ 30 | RW1 | 68.49 | 31.51 | x |

−20 ≤ y ≤ 20 | RW2 | 50.86 | 49.14 | x | |

Central Africa | −15 ≤ x ≤ 25 | R_{C}1 | 72.93 | 27.07 | x |

10 ≤ y ≤ 32 | R_{C}2 | 67.21 | 32.79 | x | |

East Africa | −25 ≤ x ≤ 18 | R_{E}1 | 57.31 | 42.69 | x |

30 ≤ y ≤ 55 | R_{E}2 | 57.74 | 40.26 | x | |

Southern Africa | −40 ≤ x ≤−18 | RS1 | 54.75 | 45.25 | x |

10 ≤ y ≤ 35 | RS2 | 72.41 | 27.53 | x |

Models | Correlation coefficient |
---|---|

R_{N}_{1} | 0.954699 |

R_{N}_{2} | 0.901375 |

R_{N}_{3} | 0.990823 |

R_{N}_{4} | 0.990259 |

R_{W}_{1} | 0.992251 |

R_{W}_{2} | 0.948107 |

R_{C}_{1} | 0.991575 |

R_{C}_{2} | 0.996414 |

R_{E}_{1} | 0.924941 |

R_{E}_{2} | 0.911480 |

R_{S}_{1} | 0.911290 |

R_{S}_{2} | 0.900962 |

The resulting models for global solar radiation in the five subregions of Africa, as shown in Equations (1)-(5), indicate that each of the models is unique and in line with reality, since solar radiation is always available even during the harshest winter or most flooding rainy season.

An important aspect of empirical modelling is to build models that generalize well, and this is central to the success of a model to be fit for use as a predictor. This capacity has been demonstrated by means of the high level of predictability (with correlation coefficients not less than 0.9) achieved with the new models.

The factor analysis indicates that in determining the seasonal global solar radiation for Africa’s subregions, latitude played the major role with an overall contribution of 63.78%. The dominance of latitude as a component of the location factor buttresses some aspect of the information in [

A comparison of the global solar radiations generated and graphically illustrated in Figures 8-12 indicate that the new empirical models made very reliable predictions as evidenced by the close agreement between the predicted and actual values of global solar radiation. From

Another merit is that the domain of application, i.e. the subregions of the African continent, is the largest found in the literature.

The mean daily global solar radiation has been considered in this study as dependents of locational and seasonal factors while the data for the African continent were used for illustrating the methodology adopted. The new models obtained were presented in closed form for the five subregions of Africa.

The novelty in this paper is that accurate, reliable and computationally less burdensome empirical models with a few input factors were developed for use as predictive tools in estimating global solar radiation derivable for any location. This study also affirms the latitude as the dominant locational factor. In addition, the domain of application in the study is the largest so far.

Finally, the new models featured optimal performance in respect of estimating global solar radiation for any practical application.

The authors declare no conflicts of interest regarding the publication of this paper.

Bamigbola, O.M. and Atolagbe, S.E. (2021) Empirical Models for Predicting Global Solar Radiation on the African Continent Based on Factors of Location and Season. Open Journal of Modelling and Simulation, 9, 59-73. https://doi.org/10.4236/ojmsi.2020.91004