scholarly journals Global Solar Radiation Forecasting Using Square Root Regularization-Based Ensemble

2019 ◽  
Vol 2019 ◽  
pp. 1-20 ◽  
Author(s):  
Yao Dong ◽  
He Jiang
Keyword(s):  
Solar Radiation ◽  
Ensemble Learning ◽  
Original Data ◽  
Global Solar Radiation ◽  
Percentage Error ◽  
Square Root ◽  
Absolute Deviation ◽  
System A ◽  
Main Challenge ◽  
Accelerated Gradient

In recent decades, the integration of solar energy sources has gradually become the main challenge for global energy consumption. Therefore, it is essential to predict global solar radiation in an accurate and efficient way when estimating outputs of the solar system. Inaccurate predictions either cause load overestimation that results in increased cost or failure to gather adequate supplies. However, accurate forecasting is a challenging task because solar resources are intermittent and uncontrollable. To tackle this difficulty, several machine learning models have been established; however, the forecasting outcomes of these models are not sufficiently accurate. Therefore, in this study, we investigate ensemble learning with square root regularization and intelligent optimization to forecast hourly global solar radiation. The main structure of the proposed method is constructed based on ensemble learning with a random subspace (RS) method that divides the original data into several covariate subspaces. A novel covariate-selection method called square root smoothly clipped absolute deviation (SRSCAD) is proposed and is applied to each subspace with efficient extraction of relevant covariates. To combine the forecasts obtained using RS and SRSCAD, a firefly algorithm (FA) is used to estimate the weights assigned to individual forecasts. To handle the complexity of the proposed ensemble system, a simple and efficient algorithm is derived based on a thresholding rule and accelerated gradient method. To illustrate the validity and effectiveness of the proposed method, global solar radiation datasets of eight locations of Xinjiang province in China are considered. The experimental results show that the proposed RS-SRSCAD-FA achieves the best performances with a mean absolute percentage error, root-mean-square error, Theil inequality coefficient, and correlation coefficient of 0.066, 20.21 W/m2, 0.016, 3.40 s, and 0.98 in site 1, respectively. For the other seven datasets, RS-SRSCAD-FA still outperforms other approaches. Finally, a nonparametric Friedman test is applied to perform statistical comparisons of results over eight datasets.

scholarly journals Comparative Study of Ground Measured, Satellite-Derived, and Estimated Global Solar Radiation Data in Nigeria

2016 ◽  
Vol 2016 ◽  
pp. 1-7 ◽  
Author(s):  
Boluwaji M. Olomiyesan ◽  
Onyedi D. Oyedum
Keyword(s):  
Solar Radiation ◽  
Comparative Study ◽  
Global Radiation ◽  
Statistical Error ◽  
Global Solar Radiation ◽  
Coefficient Of Determination ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Radiation Data ◽  
Proposed Model

In this study, the performance of three global solar radiation models and the accuracy of global solar radiation data derived from three sources were compared. Twenty-two years (1984–2005) of surface meteorological data consisting of monthly mean daily sunshine duration, minimum and maximum temperatures, and global solar radiation collected from the Nigerian Meteorological (NIMET) Agency, Oshodi, Lagos, and the National Aeronautics Space Agency (NASA) for three locations in North-Western region of Nigeria were used. A new model incorporating Garcia model into Angstrom-Prescott model was proposed for estimating global radiation in Nigeria. The performances of the models used were determined by using mean bias error (MBE), mean percentage error (MPE), root mean square error (RMSE), and coefficient of determination (R2). Based on the statistical error indices, the proposed model was found to have the best accuracy with the least RMSE values (0.376 for Sokoto, 0.463 for Kaduna, and 0.449 for Kano) and highest coefficient of determination, R2 values of 0.922, 0.938, and 0.961 for Sokoto, Kano, and Kaduna, respectively. Also, the comparative study result indicates that the estimated global radiation from the proposed model has a better error range and fits the ground measured data better than the satellite-derived data.

scholarly journals Estimation of monthly global solar radiation in Buenos Aires: preliminary analysis

2014 ◽  
Vol 5 (1) ◽  
pp. 669-680
Author(s):  
Susan G. Lakkis ◽  
Mario Lavorato ◽  
Pablo O. Canziani
Keyword(s):  
Solar Radiation ◽  
Buenos Aires ◽  
Meteorological Data ◽  
Global Solar Radiation ◽  
Estimation Accuracy ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Third Order ◽  
Sunshine Hours

Six existing models and one proposed approach for estimating global solar radiation were tested in Buenos Aires using commonly measured meteorological data as temperature and sunshine hours covering the years 2010-2013. Statistical predictors as mean bias error, root mean square, mean percentage error, slope and regression coefficients were used as validation criteria. The variability explained (R2), slope and MPE indicated that the higher precision could be excepted when sunshine hours are used as predictor. The new proposed approach explained almost 99% of the RG variability with deviation of less than ± 0.1 MJm-2day-1 and with the MPE smallest value below 1 %. The well known Ångström-Prescott methods, first and third order, was also found to perform for the measured data with high accuracy (R2=0.97-0.99) but with slightly higher MBE values (0.17-0.18 MJm-2day-1). The results pointed out that the third order Ångström type correlation did not improve the estimation accuracy of solar radiation given the highest range of deviation and mean percentage error obtained.  Where the sunshine hours were not available, the formulae including temperature data might be considered as an alternative although the methods displayed larger deviation and tended to overestimate the solar radiation behavior.

A novel ensemble learning approach for hourly global solar radiation forecasting

2021 ◽  
Author(s):  
Mawloud Guermoui ◽  
Said Benkaciali ◽  
Kacem Gairaa ◽  
Kada Bouchouicha ◽  
Tayeb Boulmaiz ◽  
...  
Keyword(s):  
Solar Radiation ◽  
Ensemble Learning ◽  
Global Solar Radiation ◽  
Learning Approach

scholarly journals An evaluation of the artificial neural network based on the estimation of daily average global solar radiation in the city of Surabaya

2021 ◽  
Vol 22 (3) ◽  
pp. 1245
Author(s):  
Adi Kurniawan ◽  
Anisa Harumwidiah
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Solar Radiation ◽  
Global Solar Radiation ◽  
Percentage Error ◽  
Average Percentage ◽  
Daily Average ◽  
Artificial Neural ◽  
Average Percentage Error ◽  
The City

The estimation of the daily average global solar radiation is important since it increases the cost efficiency of solar power plant, especially in developing countries. Therefore, this study aims at developing a multi layer perceptron artificial neural network (ANN) to estimate the solar radiation in the city of Surabaya. To guide the study, seven (7) available meteorological parameters and the number of the month was applied as the input of network. The ANN was trained using five-years data of 2011-2015. Furthermore, the model was validated by calculating the mean average percentage error (MAPE) of the estimation for the years of 2016-2019. The results confirm that the aforementioned model is feasible to generate the estimation of daily average global solar radiation in Surabaya, indicated by MAPE of less than 15% for all testing years.

Evaluation of the regression parameters of the Angstrom-Page model for predicting global solar radiation

2013 ◽  
Vol 24 (2) ◽  
pp. 46-49 ◽  
Author(s):  
Solomon Agbo
Keyword(s):  
Solar Radiation ◽  
Statistical Tests ◽  
Global Solar Radiation ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Statistical Parameters ◽  
New Model ◽  
Page Model ◽  
Regression Parameters

A simple and empirical model for the estimation of average monthly global solar radiation for a Nigerian location is presented. Regression coefficients satisfying the Angstrom-page model have been obtained using clearness index (KT) and the relative sunshine data for the location. The test of validity of the model was done by evaluating the following statistical parameters: the mean bias error (MBE), root mean square error (RMSE), mean percentage error (MPE) and the correlation coefficient (CC). The results obtained from the statistical tests show that the new model is reliable for high precision estimation of global solar radiation. A comparison between the new model and other models is presented.

scholarly journals A Comparative Study of Some Regression Models to Estimate the Global Solar Radiation on a Horizontal Surface from Sunshine Duration and Meteorological Parameters for Ghardaïa Site, Algeria

2013 ◽  
Vol 2013 ◽  
pp. 1-11 ◽  
Author(s):  
Kacem Gairaa ◽  
Yahia Bakelli
Keyword(s):  
Solar Radiation ◽  
Horizontal Plane ◽  
Regression Models ◽  
Meteorological Parameters ◽  
Sunshine Duration ◽  
Global Solar Radiation ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Quadratic Models

A comparison between some regression correlations for predicting the global solar radiation received on a horizontal plane has been processed. Seven models for estimating the global solar radiation from sunshine duration and two meteorological parameters (air temperature and relative humidity) are presented. The root mean square error (RMSE), mean bias error (MBE), correlation coefficient (CC), and percentage error () have been also computed to test the accuracy of the proposed models. Comparisons between the measured and the calculated values have been made. The results obtained show that the linear and quadratic models are the most suitable for estimating the global solar radiation from sunshine duration, and for the models based on meteorological parameters, Abdalla and Ojosu's models give the best performance with a CC of 0.898 and 0.892, respectively.

Evaluation of Global Solar Radiation using Sunshine Hour, Temperature and Relative Humidity at Low Land Region of Nepal

2020 ◽  
Vol 6 (1) ◽  
pp. 16-24
Author(s):  
U. Joshi ◽  
K.N. Poudyal ◽  
I.B. Karki ◽  
N.P. Chapagain
Keyword(s):  
Solar Radiation ◽  
Meteorological Data ◽  
Global Solar Radiation ◽  
Coefficient Of Determination ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Energy Potential ◽  
To Come ◽  
Empirical Constants

The accurate knowledge of solar energy potential is essential for agricultural scientists, energy engineers, architects and hydrologists for relevant applications in concerned fields. It is cleanest and freely available renewable energy measured using CMP6 Pyranometer. However, it is quite challenging to acquire accurate solar radiation data in different locations of Nepal because of the high cost of instruments and maintenances. In these circumstances, it is essential to select an appropriate empirical model to predict global solar radiation for the use of future at low land, Nepalgunj (28.102°N, 81.668°E and alt. 165 masl) for the year 2011-2012. In this paper, six different empirical models have been used based on regression technique, provided the meteorological data. The empirical constants (a = 0.61, b = 0.05, c = -0.0012 and d = -0.017) are obtained to predict Global solar radiation. The values of statistical tools such as mean percentage error, mean bias error, root mean square error, and coefficient of determination obtained for Abdalla model are 1.99%, 0.003 MJ/m2/day, 2.04 MJ/m2/day and 0.74 respectively. Using the error analysis, it is concluded that the Abdalla model is better than others. So the empirical constants of this model are utilized to predict the global solar radiation to the similar geographical sites of Nepal for the years to come and it can be used to estimate the missing data of solar radiation for the respective sites.

scholarly journals Choice of ideal sunshine hour based model to predict global solar radiation in India

MAUSAM ◽  
2021 ◽  
Vol 71 (3) ◽  
pp. 451-466
Author(s):  
SAMANTA SUMAN ◽  
BANERJEE SAON ◽  
PATRA PULAK KUMAR ◽  
MAITI SUDHANSU SEKHAR ◽  
CHATTOPADHYAY NABANSU
Keyword(s):  
Energy Source ◽  
Solar Radiation ◽  
Energy Demand ◽  
Global Solar Radiation ◽  
Exponential Model ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Energy Conversion Systems ◽  
Sunshine Hour

Solar radiation is the key energy source for most of the energy conversion systems, whether it is biological or mechanical. It is also the most fundamental energy source for future energy demand. Like most of the developing countries, India also lacks sufficient instrument facilities to measure global solar radiation (GSR) at recommended spatial interval and alternative approaches must be used to generate GSR data. In the present study, six well known empirical models were tested to estimate the GSR over twelve major cities of India using long-term global solar radiation and bright sunshine hour data. The empirical coefficients have been calculated for all the models and each location using regression analysis method. Daily GSR are then calculated using those regression constants along with statistical analysis. Results reveal that all the models shows close estimation with low mean bias error (MBE), root mean square error (RMSE) and mean percentage error (MPE) values. Among all models, linear exponential and linear logarithmic models are highly recommended for prediction of GSR throughout the country, except Shillong, where Bakircilinear exponential model is recommended. Significance tests i.e., t-test also confirms that this two model produce most significant results than others.

scholarly journals Empirical models for predicting global solar radiation using meteorological parameters for Sokoto, Nigeria

2019 ◽  
Vol 7 (2) ◽  
pp. 48
Author(s):  
Davidson O. Akpootu ◽  
Bello I. Tijjani ◽  
Usman M. Gana
Keyword(s):  
Solar Radiation ◽  
Meteorological Parameters ◽  
Meteorological Data ◽  
Climatic Variability ◽  
Global Solar Radiation ◽  
Coefficient Of Determination ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Bias Error ◽  
Radiation Maximum

The performances of sunshine, temperature and multivariate models for the estimation of global solar radiation for Sokoto (Latitude 13.020N, Longitude 05.250E and 350.8 m asl) located in the Sahelian region in Nigeria were evaluated using measured monthly average daily global solar radiation, maximum and minimum temperatures, sunshine hours, rainfall, wind speed, cloud cover and relative humidity meteorological data during the period of thirty one years (1980-2010). The comparison assessment of the models was carried out using statistical indices of coefficient of determination (R2), Mean Bias Error (MBE), Root Mean Square Error (RMSE), Mean Percentage Error (MPE), t – test, Nash – Sutcliffe Equation (NSE) and Index of Agreement (IA). For the sunshine based models, a total of ten (10) models were developed, nine (9) existing and one author’s sunshine based model. For the temperature based models, a total of four (4) models were developed, three (3) existing and one author’s temperature based model. The results of the existing and newly developed author’s sunshine and temperature based models were compared and the best empirical model was identified and recommended. The results indicated that the author’s quadratic sunshine based model involving the latitude and the exponent temperature based models are found more suitable for global solar radiation estimation in Sokoto. The evaluated existing Ångström type sunshine based model for the location was compared with those available in literature from other studies and was found more suitable for estimating global solar radiation. Comparing the most suitable sunshine and temperature based models revealed that the temperature based models is more appropriate in the location. The developed multivariate regression models are found suitable as evaluation depends on the available combination of the meteorological parameters based on two to six variable correlations. The recommended models are found suitable for estimating global solar radiation in Sokoto and regions with similar climatic information with higher accuracy and climatic variability.   

scholarly journals Prediction of daily global solar radiation using different empirical models on the basis of meteorological parameters at Trans Himalaya Region, Nepal

BIBECHANA ◽  
2021 ◽  
Vol 18 (1) ◽  
pp. 159-169
Author(s):  
Usha Joshi ◽  
I B Karki ◽  
N P Chapagain ◽  
K N Poudyal
Keyword(s):  
Relative Humidity ◽  
Solar Radiation ◽  
Meteorological Parameters ◽  
Energy Resource ◽  
Global Solar Radiation ◽  
Empirical Models ◽  
Mean Bias Error ◽  
Percentage Error ◽  
Sunshine Hour ◽  
Empirical Constants

Global Solar Radiation (GSR) is the cleanest and freely available energy resource on the earth.  GSR  was measured for six years (2010 -2015) at the horizontal surface using calibrated first-class CMP6 pyranometer at Kathmandu (Lat. 27.70o N, Long. 85.5oE and Alt. 1350m). This paper explains the daily, monthly, and seasonal variations of GSR and also compares with sunshine hour, ambient temperature, relative humidity, and precipitation to GSR. The annual average global solar radiation is about 4.16 kWh/m2/day which is a significant amount to promote solar active and passive energy technologies at the Trans-Himalaya region. In this study, the meteorological parameters are utilized in the regression technique for four different empirical models and finally, the empirical constants are found. Thus obtained coefficients are utilized to predict the GSR using meteorological parameters for the years to come. In addition, the predicted GSR is found to be closer to the measured value of GSR. The values are justified by using statistical tools such as coefficient of determination (R2), root mean square error (RMSE), mean percentage error (MPE), and mean bias error (MBE). Finally, the values of R2, RMSE, MPE, and MBE are found to be 0.792, 1.405, -1.014, and 0.011, respectively for the model (D), which are based on sunshine hour, temperature and relative humidity. In this model, the empirical constants, a = 0.155, b = 0.134, c = 0.014 and d = 0.0007 are determined which can be utilized at the similar geographical locations of Nepal. BIBECHANA 18 (2021) 159-169

Sign in / Sign up

Export Citation Format

Share Document