The potential use of ANN in this study is to validate the collected GlobalSolarRadiation (GSR) data with results obtained from ANN for prediction of solarglobalradiation for our stations and any other locations where solarenergy is required. The written computer codes were used to carry out the analysis and the simulation carried out using MATLAB software. The ANN model indicates good training performance with RMSE value of 0.0371 MJ/m 2 /day and standard deviation of 1.955 x 10 -4 MJ /m 2 /day and R 2 greater than 0.7.
Many empirical and statistical models have been developed by many researchers for forecasting GSR in Nigeria –. Artificial intelligence methods have a greater ability to handle nonlinearity and complex relations between the GSR and other meteorological data, and provide better efficiency and accuracy. Meanwhile, some researchers in many developed countries have found the techniques to be accurate alternatives to direct measurement –. Specifically, nonlinear autoregressive recurrent neuralnetworks with exogenous input (NARX) was used in  to estimate the GSR across New Zealand using air pressure, rain amount, temperature, relative humidity, azimuth angle, solar zenith angle, wind speed and wind direction. The predicted values of hourly globalsolarradiation compared favourably with the measured values. The NARX model was also used in  to predict solarradiation over India. The study used temperature, sunshine hour, and humidity as input variables and found that goodness-of-fit is 0.6431. In  an enhanced estimation of solarradiationusing NARX models with corrected input vectors was studied over three locations in Mexico. The study used wind speed, pressure, relative humidity and temperature as input vectors and the performance evaluation showed the coefficient of determination (R) of 0.947 for Chihuahua, 0.968 for Temixco and 0.957 for Zacatecas stations in Mexico. However, the artificialneural network techniques for the estimation of GSR have not been fully utilised in Nigeria , . Among the few researchers in Nigeria who have used the NARX model for estimation of solarradiation, include  that introduced a hybrid SARIMA- NARX neural network model for solarradiationestimation in Makurdi, Nigeria. The study used minimum and maximum temperatures as input variables and concluded the model has good accuracy after validation with R of 0.771. Also,  predicted solarradiationusing the NARX model over six locations in Nigeria. The
Most regions of Egypt obtain enormous amount of solarenergy due to their valuable geo- graphical place. The data used in this study are the globalsolarradiation (GSR), maximum temperature (T max), minimum temperature (T min), averages temperature (T avg), relative humidity (RH), and atmospheric pressure (Atm.p) of three different locations which are Borg El-Arab, Cairo, and Aswan. These locations varied in climatic condition across Egypt and data were collected for a period of 14 years from 1 January 2002 to 31 December 2015 which we obtained from the NASA Surface meteorology and SolarEnergy web site 1 .
The main objective of this work is to predict average daily globalsolarradiation (GSR) without any measuring instruments , in future time domain for Madurai city, located in Tamilnadu (India) by using Standard multilayered feed-forward, back-propagation neural network with Levenberg- Marquardt (LM) training algorithm and Gradient descent back propagation (GD) algorithm. In order to train and test the neural network, three different artificialneural network models are developed, based on daily average meteorological data like maximum ambient air temperature, minimum ambient air temperature and minimum relative humidity values for predicting globalsolarradiation. The measured data were randomly selected for training, validation and testing the neural network. The results from the three artificialneural network models shows that using the minimum air temperature and day of the year outperforms the other cases with absolute mean percentage error of 5.36% and mean square error of 0.006 when training was done by using LM back propagation learning algorithm. From the results it is very clear that neural network is well capable of estimating GSR from simple and available meteorological data. It is expected that the models developed for daily globalsolarradiation will be useful to the designers of energy-related systems as well as to those who need to estimate the daily variation of globalsolarradiation for the specific location in Tamilnadu (India).
developed a new formula, based on meteorological and geographical data to determine the solar-energy potential in Turkey usingartificialneuralnetworks. Scaled conjugate gradient and Levenberg–Marquardt (LM) learning algorithms and a logistic sigmoid transfer function were used in the network. Meteorological data for four years (2000–2003) from 18 cities spread over Turkey were used as training data of the neural network, Meteorological and geographical data (latitude, longitude, altitude, month, mean sunshine duration, and mean temperature) were used in the input layer of the network. Solarradiation was the output parameter as has been depicted (Sozen, Arcaklioglu, Ozalpand Kanit, 2005). Artificialneuralnetworks to develop prediction models for daily globalsolarradiationusing measured sunshine duration for 40 cities covering nine major thermal climatic zones and sub-zones in China were used (Lam, Wan and Yang, 2008). Coefficients of determination (R 2 ) for all the 40 cities and nine climatic zones/sub-zones are 0.82 or higher, indicating reasonably strong correlation between daily solarradiation and the corresponding sunshine hours. The measured air temperature and relative humidity values between 1998 and 2002 for Abha city in Saudi Arabia for the estimation of global
The use of satellite-based data overcomes the limitations of site measurements and provides an alternative for obtaining the spatial distribution of solarradiation (Ibrahim et al., 2017; Quesada-Ruiz et al., 2015). Hence, for this reason in this doctoral research thesis for objective 2, 3 and 4 the satellite based atmospheric parameter (MODIS) are used to develop Artificial Intelligence based predictive GSR model. MODIS is an instrument aboard the Terra (EOS AM) and Aqua (EOS PM) satellites. Terra's orbit around the Earth is timed so that it passes from north to south across the equator in the morning, while Aqua passes south to north over the equator in the afternoon. Terra and Aqua cover the globe every 1-2 days, providing data in moderate spatial resolution (250 m at nadir) with wide swath (2330 km) and large spectral range (36 channels between 0.412 and 14.2 μm)(López et al., 2014). Forty-four data products are retrieved from the MODIS observations. Among these products, the MOD08-M3 contains approximately 800 sub-datasets describing features of the atmosphere, such as the cloud fraction, cloud optical thickness, precipitable water vapor amount, and aerosol optical thickness (Kim et al., 2010). These remotely sensed atmospheric products can be considered as an alternative predictor for Artificial Intelligence based GSR predictive model, particularly for the remote locations with no ground-based measurement infrastructure.
Although the data filling methodology is good for the location in this study, it also can be utilized for dataestimation in other geo-location with notes, for first method the assignment criteria of atmospheric transmittance using RH and ambient temperature should be adjusted to the available solarradiationdata of the area to get minimum error. To generate general criteria of atmospheric transmittance assignment using RH and ambient temperature further research is required with sufficient large amount of data for various area. For the second method, the correlation should be build based on available measurement nearest from the location to give satisfactory estimation results.
The rate of heat absorbed by a collector plays a significant role in deciding the performance of the collector. The absorbing surface of a flat plate collector is in form of a tray and the energy collected by the aperture area is distributed between the absorber base plate and walls. For solar flat plate collectors the medium to be heated is in contact with the base plate and therefore the heat transfer is mainly through the base plate. In case of solar water heaters, the wall heights are very small (~ 5cm.) whereas the collector areas are quite large (~1-2 sq.m.) hence the shading of the base plate by the walls and the fraction of heat absorbed by the walls is less. In box type solar cookers the wall height is more (~10cm.) and the base plate area is small (~ 0.25 sq.m.) hence the study of energy distribution between base plate and walls becomes important as the effect of wall height and wall inclination is more pronounced on energy absorption by the base plate. The wall height and wall inclination play a major role in determining the shading effects on the base plate which is further visualized as non-uniform heating of load.
Initial concentration of pollutant is the important parameter that affects the rate of PCD of pollutant. The experiments on PCD of meta-chlorophenolwere conducted for solutions with range of 0.1gm/l – 1 gm/l concentrations. It was found thatmeta-chlorophenolof initial concentration50 mg/l was completely degraded within2 hr by usingsolar radiations. The PCD time increased with increase in meta-chlorophenolconcentration. In this particular study catalyst concentration was kept constant 1 gm/l. The concentration ofmeta- chlorophenolhas a significant effect on the PCD. The rate was higher when the initial concentration meta- chlorophenolwas less. This can be explained as for a certain TiO 2 concentration, the amount of active centers on
Accurate solar irradiance forecasting is essential for minimizing operational costs of solar pho- tovoltaic (PV) generation as it is commonly used to predict the power output. This thesis presents and compares three different machine learning approaches of solar irradiance forecasting: Ran- dom Forest (RF), Feedforward NeuralNetworks (FNNs) and Long Short-Term Memory (LSTM) networks. Each model was tested on two different forecasts: the next hour average and the hourly day-ahead averages. The machine learning algorithms were trained and tested on data from a weather station located at Tampere University (TAU) in Tampere, Finland. Data were pre- processed before training the algorithms and the relevant features were selected. Moreover, Grid Search and Random Search techniques were used along with multiple train and validation splits to find the optimal hyperparameters for each machine learning algorithm. Persistence model is set as a baseline model for comparison while RMSE and MAE are used to quantify the prediction error. For the next hour forecast, LSTM achieved the highest accuracy in terms of RMSE (76.14 W/m 2 ), 2.1% and 1.1% better than RF and FNN respectively. Instead, FNN generally produced
Solarradiation is the energy emitted by the Sun. Solarradiation is partly absorbed, scattered and reflected by molecules, aerosols, water vapor and clouds as it passes through the atmosphere, (Rai, 2006). Measurement of solarradiation at some locations has been found to be essential in order to really assess the availability of solarenergy arriving on the earth. The sun is the chief source of solarenergy and nearly all known elements are present in the sun however, the main constituents are hydrogen and helium. These elements constitute about 80% and 19% respectively. The sun has a mass of 1.9889 x 10 30 kg and radius of 6.960 x 10 8 m. It is at a mean distance of 150 million km from the earth with a volume and density of 1.412 x 10 18 km 3 and 1.622 x 10 5 kgm -3 respectively. Its core temperature is about 2 x 10 7 k while the outermost layer has an equivalent black body
The data of maximum and minimum temperature values adopted from the website of weather online limited have been used to determine the extraterrestrial solarradiation and globalsolarradiation of Abuja. The computations of various parameters were done in excel worksheet using their formulas. From the computations, the global and extraterrestrial solar radiations where gotten using Hargreaves-Samani’s model, such that the daily maximum and minimum globalsolar radiations are 29.606kwh and 12.044kwh in that order, while the daily maximum and minimum extraterrestrial solar radiations are 40.973kwh and 39.1kwh respectively. The maximum globalsolarradiation of 29.606kwh shows the increase in its value from 26.491kwh predicted by Ugwu, A. I and Ugwuanyi, J. U. in 2011, in their work in which maximum and minimum temperature data of Abuja adopted from Nigeria Meteorological Agency (NIMET) for the month of January, 2009 was used to predict globalsolarradiation of the area using Hargreaves-Samani’s model. The 3.118kwh rise in globalsolarradiation in seven years (2009-2016) as seen in this work may be attributed to global warming, and adequate measures should be put in place by environmental regulatory agencies in the country to prevent further such increase.
This paper presents several forecasting methodologies based on the application of ArtificialNeuralNetworks (ANN) and Support Vector Machines (SVM), directed to the prediction of the solar radiance intensity. The methodologies differ from each other by using different information in the training of the methods, i.e, different environmental complementary fields such as the wind speed, temperature, and humidity. Additionally, different ways of considering the data series information have been considered. Sensitivity testing has been performed on all methodologies in order to achieve the best parameterizations for the proposed approaches. Results show that the SVM approach using the exponential Radial Basis Function (eRBF) is capable of achieving the best forecasting results, and in half execution time of the ANN based approaches.
In this paper, a combination of data clustering and artificial intelligence tech- niques are used to predict incoming solarradiation on a daily basis. The data clustering technique known as Perceptually Important Points is proposed, where time-series data is grouped into clusters separated by key characteristic points, which are later used as training data for an artificialneural network. The type of network used is known as a Focused Time-Delay Neural Net- work, and an analysis of the data is performed using the Mean Absolute Per- centage Error scheme.
Solarradiation speaks to the biggest vitality stream entering the earthly environment. After reflection and ingestion in the climate, around 100,000TW hit the surface of Earth and experience change to all types of vitality utilized by people, except for atomic, geothermal, and tidal vitality. This asset is huge and compares to just about 6,000 crease the current worldwide utilization of essential vitality (13.7TW. In this way, solar based vitality has the capability of turning into a noteworthy segment of an economical vitality portfolio with obliged ozone harming substance discharges. Solarradiation is a sustainable power source asset that has been utilized by humankind in all ages. Inactive solar based advancements were at that point utilized by antiquated civic establishments for warming or potentially cooling residences and for water warming; in the Renaissance, convergence of solarradiation was broadly considered and in the nineteenth century the principal sun powered based mechanical motors were constructed. The revelation of photovoltaic impact by Becquerel in 1839 and the formation of the primary photovoltaic cell in the mid 1950s opened totally new points of view on the utilization of sun powered vitality for the creation of power.
In this study, the wavelet transforms and the ANN has been applied to estimate the daily precipitation. The meteorological data belong to the three station were investigated for this study. These are the daily mean temperature, the daily maximum temperature, the daily minimum temperature, the daily total spe- cific humidity, the daily total evaporation and the daily total precipitation. Each of the meteorological data considered as input for the ANN model was decomposed into the wavelet sub-series by Discrete Wavelet Transform (DWT). Then, ANN configura- tion is constructed with appropriate wavelet sub se- ries as input and the original precipitation time se- ries as output. So, different wavelet-ANN models were prepared for each station. Precipitation esti- mation was applied with the two different algorithms of the artificialneuralnetworks. Employment of the Feed Forward Back Propagation (FFBP) in the pre- cipitation estimation is compared with the Radial Basis Function (RBF) performances. The results were also compared with linear regression model. As a result, it was seen that the wavelet-feed forward back propagation method provided the best estima- tion performance. Results indicate that the wavelet- ANN model estimations were superior to the ones obtained by the multi linear regression model. The wavelet-ANN models have provided a good fit with the observed data, especially for the time series which have zero precipitation in the summer months. It was seen that the ANN-wavelet method provided very successful estimation performance. This study is the first application to the daily precipitation es- timations using the wavelet sub-series of the various meteorological variables in the water resources lit- erature. The wavelet-ANN method is especially con- venient in variables having non-linear dynamics such as predicting of precipitation data.
est. However, we consider this to have limited effects since (1) the vegetation growth/changes are continuous in time, (2) the NBAR product uses 16-day data but also emphasizes the specific day of interest (Schaaf, 2018). These four bands were selected because the visible and near-infrared bands included most of the vegetation information and drives the variation of SIF (Verrelst et al., 2015). We also tested us- ing all seven bands with/without the meteorological variables (temperature and vapor pressure deficit, obtained from the OCO-2 SIF lite files) to train the NN, but the improvements in training and validation were very minor (R 2 increased by less than 0.01; data not shown), and thus we decided not to use it. Since SIF is very sensitive to the incoming solar ra- diation, using cloud-free training samples can minimize the uncertainty of using cosine of the solar zenith angle as the proxy of incoming PAR. It should be noted that the training dataset may contain snow-affected samples, but these were not removed to get a more realistic prediction of SIF during winter.
Moderate Resolution Imaging Spectroradiometer (MODIS) is a passive imaging spectroradiometer, which covers the visible and infrared regions of electromagnetic radiation. Terra platform launched in 1999 and Aqua platform launched in 2002, provides comprehensive and frequent global earth imaging in 36 spectral bands between 0.41 and 14.39 mm and at variable spatial resolution with nadir footprints no more than 1 km. Furthermore, MODIS supplies a series of products for various land/ocean applications . The data that have been used in this paper are the MOD021KM, MOD03 and MOD05_L2 products. MOD03
Nowadays, to contrast negative effects of pollution, global warming and waste of energy, green energy represents a very attractive solution, especially solar en- ergy . Indeed, applications like Photovoltaic (PV) systems are changing the electrical energy production, consumption and distribution in our cities . We are witnessing the transaction of our society from centralized and hierarchical power distribution systems to distributed and cooperative ones, generally called Smart Grids. The technology introduced by this new philosophy is opening the electrical marketplace to new actors (e.g. prosumers and energy aggregators). In classic power grids, the stability is achieved by consolidated generation plants using primary and secondary reserve at large-scale . Whilst, in a Smart Grid scenario, new actors can actively contribute to load-balancing by fostering novel services for network management and stability. Demand/Response  is an ex- ample of such applications for Smart Grid management. It permits to achieve a temporary virtual power plant  by changing the energy consumption pat- terns of consumers i) to match energy produced by renewable energy systems or ii) to fulfil grid operation requirements. This process is generally done every 15 minutes. In these applications, the amount of available energy must be known in advance to optimize the production of power plants  and to match energy pro- duction with consumption. Thus, we need tools to forecast with a good accuracy of solarradiation and, consequently, solarenergy.
Multiple linear regression models were developed to estimate the monthly daily Sunshine Hours using four parameters during a period of eleven years (1997 – 2007) for Calabar, Nigeria (Latitude 5 o 16’07.6’’N); The parameters include Relative Humidity, Maximum and Minimum temperatures, Rainfall and Wind Speed. The result of the correlations shows that the four variable correlations with the highest value of R gives the best result when considering the error term Root Mean Square Error (RMSE). The model is given as S = -11.049-6.540RF- 0.534W+0.142RH+1.127T. Where RH is Relative humidity, T is the Difference in maximum and minimum temperature, RF is Rainfall, and W is wind speed. The developed model can be used in estimating Globalsolarradiation for Calabar and other locations with similar climatic conditions. Keywords: Globalsolarradiation, Sunshine hours.