Daily Global Solar Radiation Research Articles

Combined empirical and machine learning modeling method for estimation of daily global solar radiation for general meteorological observation stations

Daily global solar radiation (H) is practically significant for human production and life, especially for solar power generation. Due to the high construction and maintenance cost of solar radiation observation equipment, solar radiation measurement cannot be easily obtained at many sites. In addition, the H estimation fitting model cannot be directly trained in the absence of site-specific historical data of H. Therefore, in this study, the H estimation modeling method is proposed specifically for the sites that cannot afford to install solar radiation measurement equipment. The method includes a novel coding method based on information gain, Pearson correlation coefficient, and principal component analysis (PCA), which analyzes the nonlinear and high-dimensional correlation between meteorological factors and solar radiation to decide the correlation of adjacent sites. Based on the results of the coding method, a hybrid H estimation model combining empirical and machine learning is proposed, which takes the empirical model as the base model, and the machine learning model adaptively assigns the corresponding weights to estimate H. The case studies show the proposed hybrid model outperforms the benchmark models and indicate that the H estimation modeling method can be applied to different regions without solar radiation measurement.

Efficient daily solar radiation prediction with deep learning 4-phase convolutional neural network, dual stage stacked regression and support vector machine CNN-REGST hybrid model

Optimal utilisation of the sun's freely available energy to generate electricity requires efficient predictive models of global solar radiation (GSR). These are necessary to provide solar energy companies an early and effective market entry to support renewable energy integration into electrical grids. We propose a hybrid deep learning CNN-REGST method where a Convolutional Neural Network is integrated with a dual-stage Stacked Regression (Level-O Learner and Level-O predictor) followed by a Support Vector Machine (Level-1 Learner) with its hyperparameters optimised using the HyperOpt function to predict the daily GSR with high accuracy. Six solar energy farms in Queensland, Australia, are selected as testing sites and the predictive features from Global Climate Models and observations, derived using marine predator algorithm, are employed to build the CNN-REGST prediction model. We include a feature selection process based on meta-heuristic methods to select the optimal predictors used as inputs for the resulting CNN-REGST model. Our hybrid model is rigorously evaluated to analyze its performance over a yearlong, and all four season data. We also compare the proposed CNN-REGST model with several deep learning (i.e., CNN, Long-term Short-term Memory Network LSTM, Deep Neural Network DNN) and conventional ML approaches (Extreme Learning Machine ELM, Stacked Regression REGST, Random Forest Regression RFR, Gradient Boosting Machine GBM, Multivariate Adaptive Regression Splines MARS) using the same test datasets. The simulations carried out show that the proposed hybrid model is significantly accurate in GSR predictions compared with the deep learning and the ML models as well as a commonly used persistence model. We conclude that the CNN-REGST prediction model could be a useful scientific ploy incorporated in modern solar energy monitoring technologies to utilize a greater proportion of sustainable energy resources captured from the sun into consumer electricity for conventional-renewable hybrid energy grid systems.

ESTIMATION OF DAILY GLOBAL SOLAR IRRADIANCE FROM HIMAWARI-8 PRODUCTS OVER CHINA

Abstract. Daily global surface solar irradiance (SSI) is of great importance parameter in the surface energy balance, climate modelling, and solar energy utilization. However, it is still challenging to extend the estimated instantaneous solar radiation to the daily global SSI from either geostationary satellites or polar-orbiting satellites. In this study, a method for estimating the daily global SSI is proposed based on the Himawari-8 hourly SSI products using pixel-by-pixel Gaussian fitting to simulate the diurnal SSI variation. Compared with ground-based observations, the daily global SSI estimated by Gaussian fitting have higher accuracy under various weather conditions than by the simple accumulation or quadratic polynomials, and the coefficient of determination (R2) between the estimated and observed values exceeds 0.86. The verification results also show different estimation accuracies under different weather conditions. The root-mean-square errors (RMSEs) under clear-sky and all-sky conditions are 2.36 MJ/m2 and 3.06 MJ/m2, respectively. In addition, experimental results show that the daily global SSI in China has high spatial heterogeneity. For higherelevation areas with low cloud cover, such as the Qinghai-Tibet Plateau, Inner Mongolia, and Northwest China, the daily global SSI is higher than other areas. In contrast, the values of daily global SSI is relatively low in the eastern regions of Southwest and Northeast China.

Modeling daily global solar radiation using only temperature data: Past, development, and future

Accurate determination on global solar radiation (Rs) is essential in many disciplines and sectors. This study comprehensively reviewed 78 existing models and developed 4 new models (N1-1∼N1-4) for estimating daily Rs using only temperature data. The best model was then applied to forecast daily Rs using temperature data in weather forecast. All models were calibrated and tested using data collected from 105 radiation stations scattered across China. Daily data of temperature were also collected from 703 additional meteorological stations to classify diurnal temperature range (ΔT) zones in China. The results showed that China could be divided into five ΔT zones. The newly developed model N1-4 was the most accurate among all models for various ΔT zones, with values of coefficient of determination, index of agreement, mean bias error, mean absolute bias error, root mean square error, and relative root mean square error in the ranges of 0.660–0.737, 0.888–0.919, −0.023–0.100 MJ m−2 d−1, 2.691–3.195 MJ m−2 d−1, 3.725–4.231 MJ m−2 d−1, and 0.224–0.352, respectively. The generalized coefficients of the N1-4 model were also derived and further used for estimating daily Rs in areas with or without Rs data when there are only temperature data. The model N1-4 was successfully applied to predict daily Rs in varying ΔT zones with forecast temperature data for a lead time of 1–8 days. This new model is recommended for forecasting daily Rs in zones with high ΔT for a long lead time and in zones with low ΔT for a short lead time.

A New Model to Predict the Global Solar Radiation GSR of Souk-Ahras City

The value of the global solar radiation GSR reaching the earth is very important because it is an essential variable for different applications. Unfortunately, solar radiation measurements are not available, most of the time, in developing countries because of the lack of measurement means. Moreover, these measurements are difficult to obtain under complicated weather conditions. Thus, solar radiation evaluation models are used. In this study, a new semi-empirical model for the estimation and prediction of the global solar radiation of Souk Ahras area in Algeria is proposed. The model developed is based on meteorological data such as daily hourly temperatures and average relative humidity required from several stations and databases, over a period of four years. The values of the regression coefficients a, b and c calculated are 0.0142, -10.6206 and 57.8367 respectively. To set the modal valid, we have applied it to 10 Algerian cities and we calculated the H/H0 ratio for each site from our model. They have been then compared with the values from the (CDER). We can conclude that our new model gives a good estimate of the average daily global solar radiation (H) for the studied regions (error between 2.49% and 8.93%) and can also be used elsewhere in areas with the same climatic conditions.

Boosting solar radiation predictions with global climate models, observational predictors and hybrid deep-machine learning algorithms

This paper presents a new hybrid approach for Global Solar Radiation (GSR) prediction problems, based on deep learning approaches. Predictive models are useful ploys in solar energy industries to optimize the performance of photovoltaic power systems. Specifically, in this work we develop a new 4-phase hybrid CXGBRFR framework, which includes a deep learning Convolutional Neural Network, an Extreme Gradient Boosting with Random Forest Regression, and a Harris Hawks Optimization for initial feature selection. The proposed system has been tested in a problem of daily GSR prediction at six solar farms in Australia. Data from three global climate models (GCM) (CSIRO-BOM ACCESS1-0, MOHC Hadley-GEM2-CC and MRI MRI-CGCM3) have been considered as predictive (input) variables for the proposed approach. The variables from these GCMs contain enough information to obtain an accurate prediction of the GSR at each solar farm. The performance of the proposed approach is compared against different deep and shallows learning approaches: Deep Belief Network, Deep Neural Network, Artificial Neural Network, Extreme Learning Machine and Multivariate Auto-Regressive Spline models. We show that the proposed approach exhibits an excellent performance in GSR prediction, against all alternative approaches in all solar farms considered.

Hybrid deep CNN-SVR algorithm for solar radiation prediction problems in Queensland, Australia

This study proposes a new hybrid deep learning (DL) model, the called CSVR, for Global Solar Radiation (GSR) predictions by integrating Convolutional Neural Network (CNN) with Support Vector Regression (SVR) approach. First, the CNN algorithm is used to extract local patterns as well as common features that occur recurrently in time series data at different intervals. Then, the SVR is subsequently adopted to replace the fully connected CNN layers to predict the daily GSR time series data at six solar farms in Queensland, Australia. To develop the hybrid CSVR model, we adopt the most pertinent meteorological variables from Global Climate Model and Scientific Information for Landowners database. From a pool of Global Climate Models variables and ground-based observations, the optimal features are selected through a metaheuristic Feature Selection algorithm, an Atom Search Optimization method. The hyperparameters of the proposed CSVR are optimized by mean of the HyperOpt method, and the overall performance of the objective algorithm is benchmarked against eight alternative DL methods, and some of the other Machine Learning approaches (LSTM, DBN, RBF, BRF, MARS, WKNNR, GPML and M5TREE) methods. The results obtained shows that the proposed CSVR model can offer several predictive advantages over the alternative DL models, as well as the conventional ML models. Specifically, we note that the CSVR model recorded a root mean square error/mean absolute error ranging between ≈ 2.172–3.305 MJ m2/1.624–2.370 MJ m2 over the six tested solar farms compared to ≈ 2.514–3.879 MJ m2/1.939–2.866 MJ m2 from alternative ML and DL algorithms. Consistent with this predicted error, the correlation between the measured and the predicted GSR, including the Willmott’s, Nash-Sutcliffe’s coefficient and Legates & McCabe’s Index was relatively higher for the proposed CSVR model compared to other DL and Machine Learning methods for all of the study sites. Accordingly, this study advocates the merits of CSVR model to provide a viable alternative to accurately predict GSR for renewable energy exploitation, energy demand or other forecasting-based applications.

Prediction of daily global solar radiation and air temperature using six machine learning algorithms; a case of 27 European countries

The prediction of global solar radiation in a region is of great importance as it provides investors and politicians with more detailed knowledge about the solar resource of that region, which can be very beneficial for large-scale solar energy development. In this sense, the main objective of this study is to predict the daily global solar radiation data of 27 cities (Brussels, Paris, Lisbon, Madrid…), located in 27 countries, which have mostly different solar radiation distributions in Europe. In this research, six different machine-learning algorithms (Linear model (LM), Decision Tree (DT), Support Vector Machine (SVM), Deep Learning (DL), Random Forest (RF) and Gradient Boosted Trees (GBT)) are used. In the training of these algorithms, daily air temperature(Ta), wind speed(Va), relative humidity(RH) and solar radiation of these cities are used. The data is supplied from the Meteonorm tool and cover the last years grouped in two periods (1960–1990; 2000–2019). To decide on the success of these algorithms, four different statistical metrics (Average Relative Error (ARE), Average absolute Error (AAE), Root Mean Squared Error (RMSE), and R2 (R-Squared)) are discussed in the study. In addition, the forecasting of air temperature and global solar radiation of these cities in 2050 and 2100 were made using three of the most recent Intergovernmental Panel on Climate Change (IPCC) scenarios (RCP2.6; RCP 4.5, and RCP 8.5). The results show that ARE, R,2 and RMSE values of all algorithms are ranging from 0.114 to 6.321, from 0.382 to 0.985, from 0.145 to 2.126 MJ/m2, respectively. By analysing all the algorithms, it is noticed that the Decision tree exhibited the worst result in terms of R,2 and RMSE metrics. Among the six prediction algorithms, the DL was recognized as the only algorithm that exceeded the t-critical value (The t-critical value is the cutoff between retaining or rejecting the null hypothesis). Globally, all the six machine learning algorithms used in this research can be applied to predict the daily global solar radiation data with good accuracy. Despite this, the SVM model is the best model among all the six models used. It is followed by the DL, LM, GB, RF and DT, respectively.

Improvement and evaluation of daily global solar radiation decomposition models using meteorological parameters: A case study for Turkey

ABSTRACT In this study, some of the commonly used daily global solar radiation decomposition models are improved by considering ambient temperature, relative humidity, and air pressure to estimate hourly global solar irradiance from measured daily global solar radiation in Eskişehir, Turkey. Accordingly, the identified functions of the considered parameters are integrated into the selected daily global solar radiation decomposition models. The accuracy of the improved models is evaluated based on the measured monthly average hourly global solar irradiance values. In addition, the monthly average power values provided by the original and improved models are compared with the measured power output values. The results indicate that the improved models have significantly higher performance in the accuracy of both global solar radiation estimation and power output forecasting in a specific time interval. Therefore, the proposed methodology can be used to improve any global solar radiation decomposition model based on locally measured data for different regions.

Estimation of Daily Ground-Received Global Solar Radiation Using Air Pollutant Data.

Ground-received solar radiation is affected by several meteorological and air pollution factors. Previous studies have mainly focused on the effects of meteorological factors on solar radiation, but research on the influence of air pollutants is limited. Therefore, this study aimed to analyse the effects of air pollution characteristics on solar radiation. Meteorological data, air quality index (AQI) data, and data on the concentrations of six air pollutants (O3, CO, SO2, PM10, PM2.5, and NO2) in nine cities in China were considered for analysis. A city model (model-C) based on the data of each city and a unified model (model-U) based on national data were established, and the key pollutants under these conditions were identified. Correlation analysis was performed between each pollutant and the daily global solar radiation. The correlation between O3 and daily global solar radiation was the highest (r = 0.575), while that between SO2 and daily global solar radiation was the lowest. Further, AQI and solar radiation were negatively correlated, while some pollution components (e.g., O3) were positively correlated with the daily global solar radiation. Different key pollutants affected the solar radiation in each city. In Shenyang and Guangzhou, the driving effect of particles on the daily global solar radiation was stronger than that of pollutants. However, there were no key pollutants that affect solar radiation in Shanghai. Furthermore, the prediction performance of model-U was not as good as that of model-C. The model-U showed a good performance for Urumqi (R2 = 0.803), while the difference between the two models was not particularly significant in other areas. This study provides significant insights to improve the accuracy of regional solar radiation prediction and fill the gap regarding the absence of long-term solar radiation monitoring data in some areas.

ASSESSMENT OF SOLAR ENERGY POTENTIAL OF NIGERIA FOR SOLAR POWER GENERATION THROUGH NUMERICAL SOLAR IRRADIANCE MAPPING

Any thorough analysis of solar energy applications in the country requires a solar resources map. Many maps have been developed, but they only show distributions presented in colour intensities. Solar irradiance can only be estimated from such maps. Using on the spot experimental data for solar map production does not provide accurate maps because various climatic and environmental conditions such as dust haze and cloud cover affect the amount of solar energy that reaches the ground in various locations in Nigeria. This is particularly true during the harmattan period. Therefore data generated from such methods cannot be used to develop accurate numerical solar maps. Data for such applications should be collected from outer space and over a long time for various locations. The National Aeronautics and Space Administration (NASA) provides a resources base for such data. The data used in this study covered 25 years (1994-2018). The average daily global solar irradiance for various longitude-latitude nodes is calculated from such data. With these data available, it is possible to estimate the average daily global solar irradiance for any location provided the longitude and latitude is known. Irradiance values for 300 locations in the country are evaluated from details presented by the maps. To validate the map, experimental irradiance values for selected locations are compared with values extracted from the solar irradiance map. Calculated irradiance values are found to agree satisfactorily with experimental values. The maximum percentage difference between calculated and experimental values is 6.8%. This proved that the solar map developed is a satisfactory provides a satisfactory representation of solar data for the various locations and can be used for other locations where such data does not exist. KEYWORDS: Solar Energy, Solar Maps, Irradiance, Energy Patterns,

ASSESSMENT OF SOLAR ENERGY POTENTIAL OF NIGERIA FOR SOLAR POWER GENERATION THROUGH NUMERICAL SOLAR IRRADIANCE MAPPING

Any thorough analysis of solar energy applications in the country requires a solar resources map. Many maps have been developed, but they only show distributions presented in colour intensities. Solar irradiance can only be estimated from such maps. Using on the spot experimental data for solar map production does not provide accurate maps because various climatic and environmental conditions such as dust haze and cloud cover affect the amount of solar energy that reaches the ground in various locations in Nigeria. This is particularly true during the harmattan period. Therefore data generated from such methods cannot be used to develop accurate numerical solar maps. Data for such applications should be collected from outer space and over a long time for various locations. The National Aeronautics and Space Administration (NASA) provides a resources base for such data. The data used in this study covered 25 years (1994-2018). The average daily global solar irradiance for various longitude-latitude nodes is calculated from such data. With these data available, it is possible to estimate the average daily global solar irradiance for any location provided the longitude and latitude is known. Irradiance values for 300 locations in the country are evaluated from details presented by the maps. To validate the map, experimental irradiance values for selected locations are compared with values extracted from the solar irradiance map. Calculated irradiance values are found to agree satisfactorily with experimental values. The maximum percentage difference between calculated and experimental values is 6.8%. This proved that the solar map developed is a satisfactory provides a satisfactory representation of solar data for the various locations and can be used for other locations where such data does not exist. KEYWORDS: Solar Energy, Solar Maps, Irradiance, Energy Patterns,

Generalized Extreme Gradient Boosting model for predicting daily global solar radiation for locations without historical data

Information on global solar radiation (Rs) is indispensable in many fields. However, reliable measurements of Rs are challenging worldwide because of high costs and technical complexities. Here, temperature- and sunshine-based generalized Extreme Gradient Boosting (XGBoost) models were proposed to estimate daily Rs for locations where historical Rs data are unknown. Four combinations of input variables were assessed. The first two included: (1) maximum, minimum, mean, and diurnal temperature, and extra-terrestrial radiation (Ra); and (2) sunshine duration, maximum possible sunshine duration, and Ra. In the first two inputs, the latter two further included geographical variables, i.e., latitude, longitude, and altitude. The developed models were also compared with temperature- and sunshine-based generalized empirical models. Daily data of Rs, maximum and minimum temperature, and actual sunshine duration during the period of 2007–2016 from 96 radiation stations of China were collected to develop and evaluate the models. The results showed that accuracy of the generalized XGBoost models was improved when geographical variables were further included in various climate zones. The generalized XGBoost model using temperature and geographical data as inputs slightly reduced accuracy compared to the temperature-based local-trained XGBoost model but is still superior to the temperature-based generalized empirical model. Somewhat surprisingly, there was comparable performance between the generalized XGBoost model using sunshine and geographical data as inputs and the local-trained sunshine-based XGBoost model. Therefore, the generalized XGBoost model was highly recommended to estimate daily Rs incorporating sunshine/temperature data and routinely available geographical information for locations where historical data are prior unknown.

Evaluation of machine learning models for predicting daily global and diffuse solar radiation under different weather/pollution conditions

Due to the rapid development of solar energy and photovoltaic industries in China, it is crucial to provide the reliable and accurate solar radiation predictions. In this work, three commonly used machine learning models for predicting global and diffuse solar radiation were assessed in eight Chinese cities, representing different geoclimatic and pollutant conditions. According to the results; regarding the nRMSE, nMAE, nMBE and R values, coastal locations (such as Shanghai, Guangzhou, etc.) obtained higher values than inland locations (such as Lanzhou and Wuhan). Moreover, the SVM (support vector machine) highly outperformed the other models in all locations, regardless of whether the study area was arid, semiarid, semihumid or humid, followed by GLMNET (generalized linear modeling) and RF (random forest). In addition, when assessing the SVM in different locations under different climatic and pollution conditions, it was indicated that the accuracy of solar radiation prediction was closely related to the weather and pollution condition levels. In general, the global solar radiation prediction error was in line with the weather condition levels. The prediction error increased as the weather level increased. However, the relationship between the pollution condition levels and the global solar radiation prediction showed a non-linear relationship. Moreover, for the prediction results of diffuse solar radiation, its variation law with different weather and pollution condition levels was almost different from that of global solar radiation. The maximum high error occurrence probability of global solar radiation and diffuse solar radiation appeared at pollution levels 5 and 1, respectively. Overall, the SVM model demonstrated its reliability in radiation prediction under slight pollution and stable weather conditions. This is crucial in locations with scarce meteorological data and can be used to optimize the selection of geographical locations for photovoltaic power station construction.

Machine Learning Models Based on Random Forest Feature Selection and Bayesian Optimization for Predicting Daily Global Solar Radiation

Prediction of daily global solar radiation with simple and highly accurate models would be beneficial for solar energy conversion systems. In this paper, we proposed a hybrid machine learning methodology integrating two feature selection methods and a Bayesian optimization algorithm to predict H in the city of Fez, Morocco. First, we identified the most significant predictors using two Random Forest methods of feature importance: Mean Decrease in Impurity (MDI) and Mean Decrease in Accuracy (MDA). Then, based on the feature selection results, ten models were developed and compared: (1) five standalone machine learning (ML) models including Classification and Regression Trees (CART), Random Forests (RF), Bagged Trees Regression (BTR), Support Vector Regression (SVR), and Multi-Layer Perceptron (MLP); and (2) the same models tuned by the Bayesian optimization (BO) algorithm: CART-BO, RF-BO, BTR-BO, SVR-BO, and MLP-BO. Both MDI and MDA techniques revealed that extraterrestrial solar radiation and sunshine duration fraction were the most influential features. The BO approach improved the predictive accuracy of MLP, CART, SVR, and BTR models and prevented the CART model from overfitting. The best improvements were obtained using the MLP model, where RMSE and MAE were reduced by 17.6% and 17.2%, respectively. Among the studied models, the SVR-BO algorithm provided the best trade-off between prediction accuracy (RMSE=0.4473kWh/m²/day, MAE=0.3381kWh/m²/day, and R²=0.9465), stability (with a 0.0033kWh/m²/day increase in RMSE), and computational cost.

A Comparative Study of Regression Models and Meteorological Parameters to Estimate the Global Solar Radiation on a Horizontal Surface for Baghdad City, Iraq

In this study, data of the monthly average of daily global solar radiation falling on a horizontal surface, relative humidity, maximum temperature, and duration of sunshine for the city of Baghdad were collected through two sources. First, from the Iraqi Meteorological Authority (IMA) for a period extending from 1961 to 2016. The second is from NASA, for the period from 1984 to 2004. Then, four linear regression models, two single and two polynomials were formulated to calculate the values of the monthly average of daily global horizontal solar radiation (GHSR) incidents. The models calculated the monthly average of daily extraterrestrial radiation and day length, using some data provided by NASA and the IMA. To ensure the validity of the used models, a statistical test was performed for the performance of the proposed models, using the indicators mean bias Error (MBE), root mean square error (RMSE) as well as mean percentage error (MPE). The validation shows the relationship between the measured and computed values (through the analysis of the results), where a great convergence was found between the measured and calculated values. This means that the proposed models can be adapted to predict global solar radiation. The highest values of measured solar radiation were during the month of June, which were 28.555 and 27.280 MJ/m2/day from the IMA and NASA, respectively. The same applies to the radiation calculated using the four empirical models. The month of June was the highest in terms of solar radiation values. The radiation values were 28.947, 26.315, 29.699, and 26.716 MJ/m2/day for the first, second, third, and fourth models, respectively. The lowest values of measured and calculated radiation were during the month of December. Always, radiation measured by the IMA was greater than those of NASA, as well as the values of radiation calculated in the two IMA-based models were greater than the other two NSA-based models. In the absence of a method for measuring the diffuse and direct (beam) solar radiations, as well as the lack of such values by meteorological authorities, and its paramount importance, they were reported to mathematically calculate them in this study. The values of statistical indicators RMSE; MJ/m2/day, MBE; MJ/m2/day and MPE% were (0.4769, 0.0164, 0.2207), (0.8641, 0.1773, -0.9680), (0.6420, 0.3996, -1.1487), (0.9604, 0.218, -1.0225) for the first, second, third and fourth models, respectively. According to the results of the statistical test, it can be indicated that the single linear regression model, based on the IMA’s data (model No.1), is the most accurate to calculate global solar radiation for Baghdad City.

Stacked LSTM Sequence-to-Sequence Autoencoder with Feature Selection for Daily Solar Radiation Prediction: A Review and New Modeling Results

We review the latest modeling techniques and propose new hybrid SAELSTM framework based on Deep Learning (DL) to construct prediction intervals for daily Global Solar Radiation (GSR) using the Manta Ray Foraging Optimization (MRFO) feature selection to select model parameters. Features are employed as potential inputs for Long Short-Term Memory and a seq2seq SAELSTM autoencoder Deep Learning (DL) system in the final GSR prediction. Six solar energy farms in Queensland, Australia are considered to evaluate the method with predictors from Global Climate Models and ground-based observation. Comparisons are carried out among DL models (i.e., Deep Neural Network) and conventional Machine Learning algorithms (i.e., Gradient Boosting Regression, Random Forest Regression, Extremely Randomized Trees, and Adaptive Boosting Regression). The hyperparameters are deduced with grid search, and simulations demonstrate that the DL hybrid SAELSTM model is accurate compared with the other models as well as the persistence methods. The SAELSTM model obtains quality solar energy prediction intervals with high coverage probability and low interval errors. The review and new modelling results utilising an autoencoder deep learning method show that our approach is acceptable to predict solar radiation, and therefore is useful in solar energy monitoring systems to capture the stochastic variations in solar power generation due to cloud cover, aerosols, ozone changes, and other atmospheric attenuation factors.

A Time Series Forecasting of Global Horizontal Irradiance on Geographical Data of Najran Saudi Arabia

Environment-friendly and renewable energy resources are the need of each developed and undeveloped country. Solar energy is one of them, thus accurate forecasting of it can be useful for electricity supply companies. This research focuses on analyzing the daily global solar radiation (GSR) data of Najran province located in Saudi Arabia and proposed a model for the prediction of global horizontal irradiance (GHI). The weather data is collected from Najran University. After inspecting the data, I we found the dependent and independent variables for calculating the GHI. A dataset model has been trained by creating tensor of variables belonging to air, wind, peak wind, relative humidity, and barometric pressure. Furthermore, six machine learning algorithms convolutional neural networks (CNN), K-nearest neighbors (KNN), support vector machines (SVM), logistic regression (LR), random forest classifier (RFC), and support vector classifier (SVC) techniques are used on dataset model to predict the GHI. The evaluation metrics determination coefficients (R2), root mean square error (RMSE), relative root mean square error (rRMSE), mean bias error (MBE), mean absolute bias error (MABE), mean absolute percentage error (MAPE), and T-statistic (t-stat) are used for the result verification of proposed models. Finally, the current work reports that all methods examined in this work may be utilized to accurately predict GHI; however, the SVC technique is the most suitable method amongst all techniques by claiming the precise results using the evaluation metrics.

Quantitative effects of air pollution on regional daily global and diffuse solar radiation under clear sky conditions

Air pollution is one of the most important factors affecting the solar radiation reaching the earth’s surface under the clear (cloudless) sky conditions. In this work, six typical cities with serious pollution located in the north, northeast, eastern, central, southwest, and southern regions of China: Beijing, Shenyang, Shanghai, Wuhan, Wenjiang and Guangzhou were selected. The influence of air pollution on global and diffuse solar radiation (Rs and Rd) under clear sky condition was quantitatively analyzed, based on the historical daily meteorological data from January 2014 to December 2020. The results showed that the highest Rs attenuation ratio of the above six cities appeared in spring, autumn, winter, summer, autumn and winter, respectively. The largest Rd increment ratio of the six cities appeared in summer, spring, autumn, summer, autumn and spring, respectively. From 2014 to 2020, the relative attenuation of Rs caused by air pollution (AQI > 100) was largest in Beijing (6.03%), followed by Shanghai (4.21%), Wenjiang (4.15%), Shenyang (3.28%) and Wuhan (3.14%), while Guangzhou (1.32%) had the smallest relative attenuation. Additionally, the Rd relative increment was largest in Beijing (16.21%), followed by Shanghai (8.78%), Wenjiang (8.13%), Shenyang (7.76%) and Guangzhou (5.96%), while Wuhan (2.57%) had the smallest relative increment.

Interval Prediction Method for Solar Radiation Based on Kernel Density Estimation and Machine Learning

Precise global solar radiation (GSR) data are indispensable to the design, planning, operation, and management of solar radiation utilization equipment. Some examples prove that the uncertainty of the prediction of solar radiation provides more value than deterministic ones in the management of power systems. This study appraises the potential of random forest (RF), V‐support vector regression (V‐SVR), and a resilient backpropagation artificial neural network (Rprop‐ANN) for daily global solar radiation (DGSR) point prediction from average relative humidity (RHU), daily average temperature (AT), and daily sunshine duration (SD). To acquire more accurate predictions of DGSR and examine the influence of historical DGSR on the performance of point prediction models, two different model inputs are considered: (1) three meteorological variables and (2) the lags of DGSR and three meteorological variables. Then, two interval prediction methods are developed by introducing the KDE to out‐of‐bag (OOB), introducing kernel density estimation (KDE) to split conformal (SC) based on the three machine learning models. The two methods for interval prediction are denoted as OOB‐KDE and SC‐KDE. The mean absolute error (MAE), mean relative error (MRE), and Kendall rank correlation (Kendall) are used to assess the point prediction models. The performance of interval prediction methods is evaluated by the prediction interval coverage probability (PICP), prediction interval normalized average width (PINAW), and coverage width criteria (CWC). The following conclusions are drawn from this study. First, the V‐SVR model performs best with the lowest mean absolute error (MAE) of 0.016 and mean relative error (MRE) of 0.001. Second, the lags of DGSR improve the prediction accuracy by about 30%. Third, the OOB‐KDE and SC‐KDE methods improved the quality of the prediction interval (PI). OOB‐KDE improved CWC by 81%, and SC‐KDE improved CWC by 99.99%. Fourth, the best interval prediction result is obtained using the SC‐KDE method using the V‐SVR model. The average difference between its PICP and prediction interval nominal coverage (PINC) is only 3% of the PINC, and its PINAW is less than 0.007.