No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
ADVANCES IN SOLAR RADIATION ESTIMATION TECHNIQUES: A COMPREHENSIVE REVIEW
ResearchGate has not been able to resolve any citations for this publication.
The increasing demand for clean energy and the global shift towards renewable sources necessitate reliable solar radiation forecasting for the effective integration of solar energy into the energy system. Reliable solar radiation forecasting has become crucial for the design, planning, and operational management of energy systems, especially in the context of ambitious greenhouse gas emission goals. This paper presents a study on the application of auto-regressive integrated moving average (ARIMA) models for the seasonal forecasting of solar radiation in different climatic conditions. The performance and prediction capacity of ARIMA models are evaluated using data from Jordan and Poland. The essence of ARIMA modeling and analysis of the use of ARIMA models both as a reference model for evaluating other approaches and as a basic forecasting model for forecasting renewable energy generation are presented. The current state of renewable energy source utilization in selected countries and the adopted transition strategies to a more sustainable energy system are investigated. ARIMA models of two time series (for monthly and hourly data) are built for two locations and a forecast is developed. The research findings demonstrate that ARIMA models are suitable for solar radiation forecasting and can contribute to the stable long-term integration of solar energy into countries’ systems. However, it is crucial to develop location-specific models due to the variability of solar radiation characteristics. This study provides insights into the use of ARIMA models for solar radiation forecasting and highlights their potential for supporting the planning and operation of energy systems.
Current solar forecast verification processes place much attention on performance comparison of a group of competing methods. However, forecast verification ought to further answer how the best method within the group performs relative to the best-possible performance which one can attain under that forecasting situation, which makes the quantification of predictability and forecast skill immediately relevant. Unfortunately, the literature on the quantification of relative performance of solar irradiance has hitherto been lacking, and very few studies have focused on the spatial distributions of predictability and forecast skill of solar irradiance. The predictability and forecast skill of an atmospheric process depend on two concepts: (1) the growth of initial error in unresolved scale of motion, and (2) the forecast performance of the standard of reference. Based upon this formalism, predictability and forecast skill of solar irradiance in the United States are quantified and mapped. Through this study, a couple of common misconceptions in regard to irradiance predictability are refuted, and the original formulation of skill score revived.
The ability of the hybrid evolutionary auto-regression integrated moving average (ARIMA) with a transformed probabilistically optimized Gumbel (GP) model (ARIMA-GP) and empirical technique (EM-GP) to predict global solar radiation (H) was investigated in Nigeria. The outcome showed that the hybrid ARIMA-GP2 (0, 0, 0) significantly outperformed the empirical and other ARIMA configuration models. Additionally, the potential for PV generation was estimated using six different PV power technology models that were fitted to Nigeria's comparable climatic conditions. It was found that, compared to other technologies, poly-crystalline silicon (p-Si) technology yielded the highest increase in solar PV output compared to other modules in both moderate forcing-case scenarios (SSP245) and strong forcing-case scenarios (SSP585) between 2015-2050 and 2051-2099. However, in the mitigation-case scenario (SSP126), amorphous silicon (a-Si) technology produces less than a 1% increase in solar PV production, while other technologies produce less than a 1% decrease in solar PV output. The CMIP6 climate model was also used to assess the effects of climate change on global solar radiation (insolation). With the exception of the far-future sequencing period (2050-2099), as the impact of climate change intensifies, there may be a corresponding increase in insolation in the moderate forcing-and strong forcing-case scenarios in all seasons between 2015-2050 and 2051-2099, as well as in the annual resolution. However, under the mitigation-case scenario (SSP126), with the exception of the DJF season during the 2015-2050 and 2051-2099 sequencing periods, impacts of climate change resulted in a maximum decline in insolation of 2.988% in Nigeria. This suggests that solar energy should be Nigeria's primary source of renewable energy and low-carbon economic planning, but Nigeria may not achieve the net-zero energy transition by 2050 if prompt climate mitigation and adaptation measures are not implemented.
The adequate modeling and estimation of solar radiation plays a vital role in designing solar energy applications. In fact, unnecessary environmental changes result in several problems with the components of solar photovoltaic and affects the energy generation network. Various computational algorithms have been developed over the past decades to improve the efficiency of predicting solar radiation with various input characteristics. This research provides five approaches for forecasting daily global solar radiation (GSR) in two Moroccan cities, Tetouan and Tangier. In this regard, autoregressive integrated moving average (ARIMA), autoregressive moving average (ARMA), feed forward back propagation neural networks (FFBP), hybrid ARIMA-FFBP, and hybrid ARMA-FFBP were selected to compare and forecast the daily global solar radiation with different combinations of meteorological parameters. In addition, the performance in three approaches has been calculated in terms of the statistical metric correlation coefficient (R²), root means square error (RMSE), stand deviation (σ), the slope of best fit (SBF), legate’s coefficient of efficiency (LCE), and Wilmott’s index of agreement (WIA). The best model is selected by using the computed statistical metric, which is present, and the optimal value. The R² of the forecasted ARIMA, ARMA, FFBP, hybrid ARIMA-FFBP, and ARMA-FFBP models is varying between 0.9472% and 0.9931%. The range value of SPE is varying between 0.8435 and 0.9296. The range value of LCE is 0.8954 and 0.9696 and the range value of WIA is 0.9491 and 0.9945. The outcomes show that the hybrid ARIMA–FFBP and hybrid ARMA–FFBP techniques are more effective than other approaches due to the improved correlation coefficient (R2).
The randomness and volatility of solar irradiance pose a challenge to efficient solar energy development and utilization across the world, which increases the necessity of developing an efficient solar irradiance forecasting model. However, previous studies rarely emphasized the importance of transfer learning and nonlinear feature selection, especially for the studies associated with newly-built photovoltaic plants. Moreover, multi-step ahead forecasting research is limited, despite its significance to dispatching efficiency improvement of photovoltaic system. Aiming to address the research gaps, a novel hybrid deep learning framework (HDLF), consisting of the modules of feature selection, feature convolution, forecasting, self-attention, transfer learning, performance evaluation, and performance analysis, is newly proposed in this paper to perform multi-step ahead solar irradiation forecasting. Specifically, the HDLF is pre-trained based on the studied datasets of global horizontal irradiation in the source domains, and the abstract information obtained from pre-training is transferred to the HDLF of the target domain to enhance its performance. To validate the effectiveness of the proposed HDLF, two simulation experiments are carried out based on the datasets from California, USA, with a resolution of 1 h. The corresponding experimental results from the modules of performance evaluation and performance analysis indicate that the maximum improvements in mean absolute error reach 80.19% and 35.67% in two experiments, respectively, thereby confirming the superiority and feasibility of the HDLF.
The share of solar power in the global and local energy mixes has increased dramatically in the past decade. Consequently, there has been a significant rise in the interest for solar power forecasting, for different time horizons, ranging from few minutes to seasons. For day-ahead forecasts, combination of Numerical Weather Prediction (NWP) models and post-processing algorithms is the most popular approach. Many recent publications have proposed innovative NWP post-processing methods. However, because different works use different datasets, metrics, and even cross-validation methods, it is rarely possible to fairly compare results across several papers.
In this work, we propose a rigorous benchmark of several solar NWP post-processing models representative of the literature. For our results to be as general as possible, the comparison is conducted with an open dataset, over 6 years and 7 locations. In addition, we propose a novel benchmarking approach, that focuses on the systematicity of the ranking of forecasting models.
Our results show that, when used in combination with proper regularization, large predictor sets are systematically beneficial to NWP post-processing methods. They also demonstrate that more complex algorithms such as neural networks and gradient boosting generally have the lowest mean square error, while support vector regression, a more parsimonious algorithm, performs better in terms of mean absolute error. Lastly, the focus given to ranking systematicity reveals that no model is better in all occasions. This means that researchers should be measured when they conclude to the superiority of a model, in particular when testing data is scarce.
Solar irradiance forecasting has been an essential topic in renewable energy generation. Forecasting is an important task because it can improve the planning and operation of photovoltaic systems, resulting in economic advantages. Traditionally, single models are employed in this task. However, issues regarding the selection of an inappropriate model, misspecification, or the presence of random fluctuations in the solar irradiance series can result in this approach underperforming. This paper proposes a heterogeneous ensemble dynamic selection model, named HetDS, to forecast solar irradiance. For each unseen test pattern, HetDS chooses the most suitable forecasting model based on a pool of seven well-known literature methods: ARIMA, support vector regression (SVR), multilayer perceptron neural network (MLP), extreme learning machine (ELM), deep belief network (DBN), random forest (RF), and gradient boosting (GB). The experimental evaluation was performed with four data sets of hourly solar irradiance measurements in Brazil. The proposed model attained an overall accuracy that is superior to the single models in terms of five well-known error metrics.
Global horizontal irradiance (GHI) forecasting at intra-day horizons of up to 12-h ahead is vital to grid integration of solar photovoltaics, but has been fundamentally difficult for all methods that do not involve numerical weather prediction (NWP), since non-NWP methods are unable to extrapolate the data to a horizon beyond a fews hours. The European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Oceanic and Atmospheric Administration (NOAA) are the most representative weather centers in Europe and America, respectively. To understand their operational impact and value to grid integration, the ECMWF’s High Resolution (HRES) model and two models from NOAA, namely, Rapid Refresh (RAP) and High-Resolution Rapid Refresh (HRRR), are validated through the Murphy–Winkler distribution-oriented verification framework, over the year 2020, at seven locations. All forecasts are retrieved at native horizontal resolutions—9 km for HRES, 13 km for RAP, and 3 km for HRRR; it depicts the “off-the-shelf” scenario if these forecasts are to be utilized by end users. Three simple linear correction methods, each being statistically optimal in its own respect, are used to post-process the raw forecasts. It was found that 1–12-h-ahead ECMWF’s HRES forecasts have a significantly lower root mean square error (14.0–33.7%) as compared to NOAA’s HRRR (19.0–53.2%) and RAP (19.2–45.9%). Even after the large biases in HRRR and RAP forecasts are removed, those post-processed versions are still inferior to the raw HRES forecasts.
This work compares the efficiency of 45 different machine learning (ML) algorithms to provide a comprehensive and most accurate model for global horizontal solar irradiance (GHSI) prediction in Eskişehir, Turkey. The dataset is provided by NASA Prediction of Worldwide Energy Resource (POWER) as satellite data that involves some characteristic weather condition variables such as temperature, precipitation, humidity etc. over 35 years. Some ML algorithms such as Extra Trees, LightGBM, HistGB, Random Forest (RF), Bagging and Decision Tree exhibit better performance among the others with commonly used statistical evaluation metrics in literature such as coefficient of determination (R²), root mean squared error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). In addition, Extra Tress regression slightly outperformed the rest of ensemble learning methods with R² of 0.99, RMSE of 8.05, MAE of 5.67, MAPE of 4%. Finally, the outcome demonstrates that the ML algorithms belonging to ensemble learning family achieved great results in GHSI prediction at specific location.
Direct normal irradiance (DNI) is often interpreted differently in ground measurements and forecasts using numerical weather prediction (NWP) models, leading to substantial bias in DNI forecasting especially under cloudy-sky conditions. To mitigate the bias, we use the Fast All-sky Radiation Model for Solar applications with DNI (FARMS-DNI) to conduct physics-based simulations of solar radiation in the circumsolar region. The DNI is quantified using the sum of the radiation along the sun direction and the scattered radiation within the circumsolar region. FARMS-DNI is coupled with the Weather Research and Forecasting model with solar extensions (WRF-Solar) to predict day-ahead DNI in 9-km pixels over an extended domain covering the contiguous United States. The DNI forecasts for the entire year of 2018 are validated using surface-based observations at the Atmospheric Radiation Measurement (ARM) and Surface Radiation Budget Network (SURFRAD) sites and a satellite-based solar radiation product developed by the procedure of the National Solar Radiation Data Base (NSRDB). The results show that the WRF-Solar coupled with the conventional Beer-Bouguer-Lambert law underestimes the overall DNI forecasts by 30%-60%, which is a joint effect of the uncertainties in computing the circumsolar radiation and forecasting cloud properties. This bias is significantly reduced by using FARMS-DNI.
High penetration of photovoltaics (PV) has been observed in the energy market over the last decade. However, its integration into electrical grids is challenging, as solar energy is highly fluctuating given its dependence on different weather variables. Consequently, short-term forecasting of solar irradiance provides a pivotal solution to ensure optimal use of the produced energy and reduce its uncertainty. This study proposes a hybrid convolutional neural network and Multilayer perceptron (CNN–MLP) model to forecast the global irradiance 15 min ahead. The model uses images from a hemispherical sky imager, time series of GHI, and weather variables collected from a ground meteorological station in Morocco. The evaluation of the proposed model under clear, mixed, and overcast days shows that the proposed model performs better than the persistence model. The root mean square error (RMSE) varies between 13.05 W/m² and 49.16 W/m² for CNN–MLP and between 45.76 W/m² and 114.19 W/m² for persistence. The coefficient of determination (R²) varies between 0.99 and 0.94 for the MLP–CNN and between 0.98 and 0.79 for persistence. The results show that the proposed model could be an appropriate choice for short-term forecasting even under cloudy conditions.
Solar forecasters have hitherto been restricting the application of numerical weather prediction (NWP) to day-ahead forecasting scenarios. With the hourly updated NWP models, such as the National Oceanic and Atmospheric Administration’s Rapid Refresh (RAP) or High-Resolution Rapid Refresh (HRRR), it is theoretically possible to utilize NWP for hour-ahead solar forecasting. Nonetheless, for NWP-based hourly forecasts to be useful, they ought to be post-processed, because such forecasts are almost always affected by the inherent bias of dynamical weather models. In this regard, Kalman filtering is herein used as a post-processing tool. Using one year of RAP and HRRR forecasts and ground-based observations made at seven geographically diverse locations in the contiguous United States, it is demonstrated that the post-processed versions of hourly updated NWP forecasts are sometimes able to attain a higher accuracy than those from classic time-series families of models, not only at long-range, but also at short-range forecast horizons. Although these improvements are marginal in terms of squared loss (about 1%–2%), since NWP models have a very different forecast-generating mechanism (solving the governing equations of motion) from that of time series methods (extrapolating data), NWP-based forecasts can be expected to be less correlated with those forecasts from time series models. Consequently, one should find this diversity profoundly rewarding during forecast combination, for the combined forecasts are able to consistently result in smaller error than the best component forecasts.
Presently, deep learning models are an alternative solution for predicting solar energy because of their accuracy. The present study reviews deep learning models for handling time-series data to predict solar irradiance and photovoltaic (PV) power. We selected three standalone models and one hybrid model for the discussion, namely, recurrent neural network (RNN), long short-term memory (LSTM), gated recurrent unit (GRU), and convolutional neural network-LSTM (CNN–LSTM). The selected models were compared based on the accuracy, input data, forecasting horizon, type of season and weather, and training time. The performance analysis shows that these models have their strengths and limitations in different conditions. Generally, for standalone models, LSTM shows the best performance regarding the root-mean-square error evaluation metric (RMSE). On the other hand, the hybrid model (CNN–LSTM) outperforms the three standalone models, although it requires longer training data time. The most significant finding is that the deep learning models of interest are more suitable for predicting solar irradiance and PV power than other conventional machine learning models. Additionally, we recommend using the relative RMSE as the representative evaluation metric to facilitate accuracy comparison between studies.
The ever-growing interest in and requirement for green energy have led to an increased focus on research related to forecasting solar irradiance recently. This study aims to develop forecast models based on deep learning (DL) methodologies and multiple-site data to predict the daily solar irradiance in two locations of India based on the daily solar radiation data obtained from NASA’s POWER project repository over 36 years (1983–2019). The forecast modeling of solar irradiance data is performed for extracting and learning the symmetry latent in data patterns and relationships by the machine learning models and utilizing it to predict future solar data. The goodness of fit and model performance are compared with rolling window evaluation using mean squared error, root-mean-square error and coefficient of determination (R2) for evaluation. The contributions of this study can be summarized as follows: (i) time series models based on deep learning methodologies were implemented to forecast the daily solar irradiance of two locations in India in consideration of the historical data collected by NASA; (ii) the models were developed on the basis of single-location univariate data as well as multiple-location data; (iii) the accuracy, performance and reliability of the model were investigated on the basis of standard performance evaluation metrics and rolling window evaluation; (iv) the feature importance of the nearby locations with respect to forecasting target location solar irradiance was analyzed and compared based on the solar irradiance data obtained from NASA over 36 years. The results indicate that the bidirectional long short-term memory (LSTM) and attention-based LSTM models can be used for forecasting daily solar irradiance data. According to the findings, the multiple-site data with solar irradiance historical data improve upon the forecast performance of single-location univariate solar data.
Ahead-of-time forecasting of the output power of power plants is essential for the stability of the electricity grid and ensuring uninterrupted service. However, forecasting renewable energy sources is difficult due to the chaotic behavior of natural energy sources. This paper presents a new approach to estimate short-term solar irradiance from sky images. The proposed algorithm extracts features from sky images and use learning-based techniques to estimate the solar irradiance. The performance of proposed machine learning (ML) algorithm is evaluated using two publicly available datasets of sky images. The datasets contain over 350,000 images for an interval of 16 years, from 2004 to 2020, with the corresponding global horizontal irradiance (GHI) of each image as the ground truth. Compared to the state-of-the-art computationally heavy algorithms proposed in the literature, our approach achieves competitive results with much less computational complexity for both nowcasting and forecasting up to 4 h ahead of time.
Optimal management for solar energy systems requires quality data to build accurate models for predicting the behavior of solar radiation. Solar irradiance and environmental data are provided by satellite and in-situ measurements. It is usual that satellite measurements present high temporal resolution with limited spatial resolution, and in-situ measurements provide high accuracy but significant missing data. This paper proposes a methodology based on machine learning algorithms that: i) takes the best of both data sources to obtain an improved spatio-temporal resolution, known as site-adaptation; and ii) provides highly accurate forecasting solar-radiation models based on deep learning on the improved data. Through a study case with real data, we show the benefits of using the proposed methodology based on machine and deep learning techniques to integrate data from different sources and to construct precise solar radiation forecasting models in regions where solar energy systems are required. Results show that machine learning models for site-adaptation performed up to 38% better than traditional methods.
Machine learning (ML) is the scientific study of algorithms and statistical models that computer systems use to perform a specific task without being explicitly programmed. Learning algorithms in many applications that's we make use of daily. Every time a web search engine like Google is used to search the internet, one of the reasons that work so well is because a learning algorithm that has learned how to rank web pages. These algorithms are used for various purposes like data mining, image processing, predictive analytics, etc. to name a few. The main advantage of using machine learning is that, once an algorithm learns what to do with data, it can do its work automatically. In this paper, a brief review and future prospect of the vast applications of machine learning algorithms has been made.
Global demand for energy is increasing rapidly and natural energy resources such as oil, gas and uranium are declining due to the widespread diffusion and development of the industry in recent years. To cover energy needs, research is being conducted on renewable energy. One of the renewable energies that can meet the world's demand so far is solar energy, which is free and inexhaustible in most parts of the world, and it has become an economic source. In this article we will make a forecast of the empirical Campbell model which will allow us to estimate the daily global irradiation on a horizontal plane and to compare it with the results measured at the Adrar site. The results show that the mean absolute percentage error (MAPE) less than 7%, the mean bias error does not exceed 3% in absolute value, relative RMSE does not exceed 7% and the correlation coefficient greater than 0.99 for the annual global radiation. It was concluded that this model could be used to predict the global solar radiation for Adrar site and for other sites with similar climatic conditions.
Solar irradiance is fluctuating and intermittent in nature. In order to optimally harness solar energy, this variability needs to be accounted for. Forecasting solar radiation proves to be helpful in optimal design, and operation of solar-energy based systems. This paper presents a solar irradiance forecasting scheme for multi-horizon forecasting of solar radiation considering 3/6/24 hours ahead scenarios. The proposed model uses long short term memory network, considering the dependence between hours of the same day along with other variables such as: direct horizontal irradiance, direct normal irradiance, relative humidity, dew point, temperature, wind speed, and wind direction. Solar radiations for four different locations of the Thar desert region have been forecasted. The model is optimized in terms of number of neurons and is evaluated using standard statistical indicators: RMSE and MAPE. RMSE for four different locations varied in the range of 0.099 to 0.181, along with MAPE values, which range from 6.79% to 10.47%. Low values of RMSE and MAPE indicate the efficacy of the proposed model.
Solar renewable energy (SRE) applications are substantial in eradicating the rising global energy shortages and reversing the approaching environmental apocalypse. Hence, effective solar irradiance forecasting models are crucial in utilizing SRE efficiently. This paper introduces a partially amended hybrid model (PAHM) by the implementation of a new algorithm. The algorithm innovatively utilizes bi-directional gated unit (Bi-GRU), autoregressive integrated moving average (ARIMA) and naive decomposition models to predict solar irradiance in 5-min and 60-min intervals. Meanwhile, the models’ generalizability strengths would be tested under an 11-fold cross-validation and are further classified according to their computational costs. The dataset consists of 32 months’ solar irradiance and weather conditions records. A fundamental result of this study was that the single models (Bi-GRU and ARIMA) outperformed the hybrid models (PAHM, classical hybrid model) in the 5-min predictions, negating the assumptions that hybrid models oust single models in every time interval. PAHM provided the highest accuracy level in the 60-min predictions and improved the accuracy levels of the classical hybrid model by 5%, on average. The single models were rigorous under the 11-fold cross-validation, performing well with different datasets; although the computational efficiency of the Bi-GRU model was, by far, the best among the models.
Accurate prediction of solar irradiance is beneficial in reducing energy waste associated with photovoltaic power plants, preventing system damage caused by the severe fluctuation of solar irradiance, and stationarizing the power output integration between different power grids. Considering the randomness and multiple dimension of weather data, a hybrid deep learning model that combines a gated recurrent unit (GRU) neural network and an attention mechanism is proposed forecasting the solar irradiance changes in four different seasons. In the first step, the Inception neural network and ResNet are designed to extract features from the original dataset. Secondly, the extracted features are inputted into the recurrent neural network (RNN) network for model training. Experimental results show that the proposed hybrid deep learning model accurately predicts solar irradiance changes in a short-term manner. In addition, the forecasting performance of the model is better than traditional deep learning models (such as long short term memory and GRU).
Microgrid is becoming an essential part of the power grid regarding reliability, economy, and environment. Renewable energies are main sources of energy in microgrids. Long-term solar generation forecasting is an important issue in microgrid planning and design from an engineering point of view. Solar generation forecasting mainly depends on solar radiation forecasting. Long-term solar radiation forecasting can also be used for estimating the degradation-rate-influenced energy potentials of photovoltaic (PV) panel. In this paper, a comparative study of different deep learning approaches is carried out for forecasting one year ahead hourly and daily solar radiation. In the proposed method, state of the art deep learning and machine learning architectures like gated recurrent units (GRUs), long short term memory (LSTM), recurrent neural network (RNN), feed forward neural network (FFNN), and support vector regression (SVR) models are compared. The proposed method uses historical solar radiation data and clear sky global horizontal irradiance (GHI). Even though all the models performed well, GRU performed relatively better compared to the other models. The proposed models are also compared with traditional state of the art methods for long-term solar radiation forecasting, i.e., random forest regression (RFR). The proposed models outperformed the traditional method, hence proving their efficiency.
Global solar radiation estimation is increasingly acquiring more importance to ensure effective management and operation of solar energy systems as well as providing reliable information about the future power generation. In this study, the Long Short-Term Memory (LSTM) has been suggested to effectively model the daily solar radiation prediction problem. The effectiveness of the suggested method compared with the state of the art machine learning algorithms such as Decision Tree Regression, Random Forest Regression, Gradient Boosting and K-Nearest Neighbor. Daily solar irradiance sequential time series data in Çorum - Turkey between January1983 and December-2018 have been used to validate the effectiveness of the suggested LSTM method. The simulation outcomes demonstrate that LSTM method has generally better performance than the other machine learning models.
Yenilenebilir enerji kaynakları sınırsız doğal döngüler sonucu oluşan temiz enerji kaynaklarıdır. Güneş enerjisi de, güneş ışınımından enerji üretimi sağlayan önemli bir yenilenebilir enerji kaynağıdır. Güneş enerjisi güç tahmininde güneşlenme şiddeti ve sıcaklık etkili faktörlerdir. Bu faktörler de hava şartlarına göre değişkenlik gösterdiğinden dolayı, güneş enerjisi santrallerinde geleceğe dair üretilebilecek enerji miktarını tahmin etmek zordur. Bu çalışmada
güneşlenme şiddeti ve sıcaklık değerleri kullanılarak bir sonraki yılda güneş panellerinden alınacak güç değerlerinin tahmini gerçekleştirilmiştir. Veri seti olarak 2013-2017 yılları arasında Niğde bölgesine ait sıcaklık ve güneşlenme şiddeti verileri kullanılmıştır. Çalışmada ilk 4 yılın verisi eğitim veri seti için, son yılın verisi ise test verisi için kullanılmıştır. Veri eğitimi için makine öğrenmesi algoritmalarından kNN, Destek Vektör Makinesi ve Lasso Regresyon algoritmaları kullanılmıştır. Sonuç olarak güneşlenme şiddeti ve sıcaklık verileri kullanılarak gelecek yıla ait güneş enerjisi potansiyeli başarılı bir şekilde tahmin edilmiştir. En iyi performansı ise 0.997 R2 skoru ile destek vektör makinesi algoritması göstermiştir.
Variation in solar irradiance causes power generation fluctuations in solar power plants. Power grid operators need accurate irradiance forecasts to manage this variability. Many factors affect irradiance, including the time of year, weather and time of day. Cloud cover is one of the most important variables that affects solar power generation, but is also characterized by a high degree of variability and uncertainty. Deep learning methods have the ability to learn long-term dependencies within sequential data. We investigate the application of Gated Recurrent Units (GRU) to forecast solar irradiance and present the results of applying multivariate GRU to forecast hourly solar irradiance in Phoenix, Arizona. We compare and evaluate the performance of GRU against Long Short-Term Memory (LSTM) using strictly historical solar irradiance data as well as the addition of exogenous weather variables and cloud cover data. Based on our results, we found that the addition of exogenous weather variables and cloud cover data in both GRU and LSTM significantly improved forecasting accuracy, performing better than univariate and statistical models.
The accurate prediction of surface solar irradiance is of great significance for the generation of photovoltaic power. Surface solar irradiance is affected by many random mutation factors, which means that there are great challenges faced in short-term prediction. In Northwest China, there are abundant solar energy resources and large desert areas, which have broad prospects for the development of photovoltaic (PV) systems. For the desert areas in Northwest China, where meteorological stations are scarce, satellite remote sensing data are extremely precious exploration data. In this paper, we present a model using FY-4A satellite images to forecast (up to 15–180 min ahead) global horizontal solar irradiance (GHI), at a 15 min temporal resolution in desert areas under different sky conditions, and compare it with the persistence model (SP). The spatial resolution of the FY-4A satellite images we used was 1 km × 1 km. Particle image velocimetry (PIV) was used to derive the cloud motion vector (CMV) field from the satellite cloud images. The accuracy of the forecast model was evaluated by the ground observed GHI data. The results showed that the normalized root mean square error (nRMSE) ranged from 18.9% to 21.6% and the normalized mean bias error (nMBE) ranged from 3.2% to 4.9% for time horizons from 15 to 180 min under all sky conditions. Compared with the SP model, the nRMSE value was reduced by about 6%, 8%, and 14% with the time horizons of 60, 120, and 180 min, respectively.
Bu çalışmada, Nisan 2017-Mart 2018 tarihleri aralığında Mersin için ölçülen günlük toplam global güneş ışınım değerlerinin yapay sinir ağı kullanılarak modellenmesi yapılmıştır ve literatürde bulunan yaygın modellerin günlük toplam global güneş ışınım değerlerini tahmin etme performansları incelenmiştir. Günlük ortalama hava sıcaklığı, bağıl nem, rüzgar hızı, güneşlenme süresi ve bulut kapalılığı verileri, Devlet Meteoroloji İşleri Genel Müdürlüğü'nden temin edilmiş olup güneş ışınım değerleri ise piranometre ile ölçülmüştür. Sonuç olarak, incelenen modeller içerisinde en iyi tahmin performansını belirlilik katsayısı (R 2) 0,83 olan Model 37 (Zhang ve Huang) göstermiştir.
In this study, the daily total global solar radiation values measured for Mersin between April 2017 and March 2018 were modeled using artificial neural networks and the performance of estimating daily total global solar radiation values of the common models in the literature was investigated. Daily average air temperature, relative humidity, wind speed, sunshine duration, and cloud cover data are obtained from the Turkish State Meteorological Service and solar radiation values are measured with a pyranometer. As a result, Model 37 (Zhang and Huang) showed the best prediction performance among the models examined, with the coefficient of determination (R 2) being 0.83.
Solar photovoltaic (PV) power generation is one of the most important sources for renewable energy. However, PV power generation is entirely dependent on the amount of downward solar radiation reaching the solar cells. This is determined by uncertain and uncontrollable meteorological factors such as temperature, humidity, wind speed, and direction, as well as other factors such as topographical characteristics. Good solar radiation prediction models can increase energy output while decreasing the operation costs of photovoltaic power generation. For example, in some provinces in China, PV stations are required to upload short-term online power forecast information to power dispatching agencies. Numerous AI, statistical, and numerical weather prediction models have been used in many real-world renewable energy applications, with a focus on modeling accuracy. However, there is a need for Explainable AI (XAI) models that could be easily understood, analyzed, and augmented by the stakeholders. In this paper, we present a compact, explainable, and lifelong learning metaheuristic-based Interval Type-2 Fuzzy Logic System (IT2FLS) for Solar Radiation Modeling. The generated model will be composed of a small number of short IF-Then rules that have been optimized via simulated annealing to produce models with high prediction accuracy. These models are updated through a life-long learning approach to maximize their accuracy and maintain interpretability. In the process of lifelong learning, the proposed method transferred the model's knowledge to new geographical locations with minimal forgetting. The proposed method achieved good prediction accuracy and outperformed on new geographical locations other transparent and black-box models by 13.2% as well as maintaining excellent generalization ability. The resulting models have been evaluated and accepted by experts, and thanks to the generated transparency, the experts were able to augment the models with their expertise, which increased the models' accuracy.
The demand for energy generation from solar energy resource has been exponentially increasing in recent years. It is integral for a grid operator to maintain the balance between the demand and supply of the grid. Solar radiation forecasting paves the way for proper planning, reserve management, and elude penalty since solar energy is sporadic in nature. Several methods can forecast solar radiation; the prior classifications are machine learning models, numerical weather prediction models, satellite imaging, sky imager and hybrid model. This article presents a comprehensive review of all those models with the working principle, challenges and future research direction. Sky imagers provide the Normalized Root Mean Square Error (nRMSE) value of 6%–9% for a time horizon of 30 min, and the satellite imagery technique provides the Root Mean Square Error (RMSE) value of 61.28 W/m² – 346.05 W/m² for a time horizon of 4 h ahead. Similarly, NWP mesoscale models provide the RMSE value of 411.6 W/m² - for three days ahead of forecasting with a spatial resolution of 50 km. Machine learning models are good at delivering accurate results with the time horizon up to 1 day ahead by yielding the results of RMSE in the range of 0.1170 W/m² – 93.04 W/m². Deep learning and hybrid models are being developed to overcome the issues faced by the standalone techniques. In many research works, artificial intelligence techniques are integrated with NWP models, sky imagers and satellite imagers to improve the data handling algorithm, which implicitly results in forecasting accuracy.
Solar radiation is one of the cleanest sources of renewable energy, and it affects the carbon sink functions of terrestrial ecosystems. Although efforts have been made to establish solar radiation observation stations around the world, their coverage remains limited. Hence, the development of a wide variety of models and techniques is indispensable for obtaining effective solar radiation data. The aim of this study is to develop hybrid models with high computational speed and high accuracy to estimate global solar radiation (GSR) and quantify the uncertainty in GSR simulations caused by uncertainty in the measurements of atmospheric and surface parameters. The radiative transfer model (RTM) library for radiative transfer (LibRadtran) was coupled with six machine learning models: extreme gradient boosting (XGBoost), random forest (RF), multivariate adaptive regression splines (MARS), multilayer perceptron (MLP), deep neural networks (DNNs), and light gradient boosting machine (LightGBM). The estimated GSR was first compared to the inversion values of the GSR provided by the Aerosol Robotic Network (AERONET) and then validated using ground-based measurements at three locations in China from 2005 to 2018. The results showed that the RTM-RF is superior in terms of computational efficiency and performance, with a mean absolute errors (MAE) and coefficients of determination (R²) of 15.57 W m⁻² and 0.98, respectively. Under clear sky conditions, aerosol optical depth (AOD) contributed the most to the accuracy of GSR estimates, with an average contribution of 57.95%. The measurement uncertainty due to the asymmetry factor, AOD, single-scattering albedo, and land surface albedo (LSA) can explain the differences in GSR between RTM estimates and GSR observations at the Lulin (20.33 vs. 20.91 W m⁻²), Wuhan (−1.40 vs. 14.58 W m⁻²), and Xianghe (7.28 vs. 14.32 W m⁻²) sites. Our study supports the use of physical models combined with machine learning models to estimate GSR and provides valuable scientific information for large-area solar radiation estimations.
Global solar radiation (GSR) prediction plays an essential role in planning, controlling and monitoring solar power systems. However, its stochastic behaviour is a significant challenge in achieving satisfactory prediction results. This study aims to design an innovative hybrid prediction model that integrates a feature selection mechanism using a Slime-Mould algorithm, a Convolutional-Neural-Network (CNN), a Long-Short-Term-Memory Neural Network (LSTM) and a final CNN with Multilayer-Perceptron output (SCLC algorithm hereafter). The proposed model was applied to six solar farms in Queensland (Australia) at daily temporal horizons in six different time steps. The comprehensive benchmarking of the obtained results with those from two Deep-Learning (CNN-LSTM, Deep-Neural-Network) and three Machine-Learning (Artificial-Neural-Network, Random-Forest, Self-Adaptive Differential-Evolutionary Extreme-Learning-Machines) models highlighted a higher performance of the proposed prediction model in all the six selected solar farms. From the results obtained, this work establishes that the designed SCLC algorithm could have a practical utility for applications in renewable and sustainable energy resource management.
The growing threat of global climate change stemming from the huge carbon footprint left behind by fossil fuels has prompted interest in exploring and utilizing renewable energy resources. Several statistical, Machine and Deep Learning techniques exist and have been used for many years for a range of forecasting problems. This study is based on the data recorded for 4 years and 9 months using precise instruments, in Islamabad, Pakistan. For this purpose we use statistical and Deep Learning architectures for forecasting solar Global Horizontal Irradiance which not only helps in grid management and power distribution, but also brings attention towards the potential of solar power production in Pakistan and its part to play in tackling global climate change. We have used statistical methods such as Seasonal Auto-Regressive Integrated Moving Average Exogenous (SARIMAX) and Prophet, and Machine Learning methods such as Long Short-Term Memory (LSTM) which is an extension of Recurrent Neural Networks (RNN), Convolutional Neural Network (CNN) and Artificial Neural Network (ANN). The selected forecast methods in our study are based on their ability to work with time series data and we have used different models configurations to see which performs best for our dataset. The performance of every model is studied using different error metrics such as Coefficient of Determination (R2), Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE). The major contribution of this study is the data collected to carry out research towards the goal of renewable energy future, and from the test methods used on the data in this study, it can be intuitively determined that ANN, CNN, and LSTM architectures perform best for short-term forecasts, while SARIMAX and Prophet are efficient for long-term forecasts.
The integrated renewable energy system is a critical component of the smart city. Integrating renewable energy sources is beneficial in addressing energy supply and demand challenges. Several methods are used to anticipate renewable energy dynamics, including assessing the current situation or projecting the future with an emphasis on a definite objective of concern. The existing degree of predictive model training includes a target country's or region's current state and projections, as well as policies and programs for future penetration. However, with the present exponential rise in renewable energy and artificial intelligence research, as well as the removal of certain existing limits, this influence may expand to include other targets in the future. Nonetheless, previous studies have omitted important features. For future research to cover more objectives, the exponential increase in the renewable energy share and rapid progress of artificial intelligence must be accompanied by necessary supervisory knowledge and technical control.
Accurate prediction of solar energy is an important issue for photovoltaic power plants to enable early participation in energy auction industries and cost-effective resource planning. This article introduces a new deep learning-based multistep ahead approach to improve the forecasting performance of global horizontal irradiance (GHI). A deep convolutional long short-term memory is used to extract optimal features for accurate prediction of the GHI. The performance of such deep neural networks directly depends on their architectures. To deal with this problem, a swarm evolutionary optimization method, called the sine-cosine algorithm, is applied and advanced to automatically optimize the network architecture. A three-phase modification model is proposed to increase the diversity of population and avoid premature convergence in the optimization mechanism. The performance of the proposed method is investigated using three datasets collected from three solar stations in the east of the United States. The experimental results demonstrate the superiority of the proposed method in comparison to other forecasting models.
After the oil crisis in 1973, the planet has to believe in alternate sources of energy aside from conventional energy resources. Among different renewable resources, solar energy is one of the most famous and attractive suppliers of electric power generation due to zero emission of CO2. However, the forecasting of a solar component in advance is one of the crucial task due to the uncertain behavior of climatic parameters. This paper provides an analysis of the techniques used in the literature to forecast solar irradiance. The study's main goal is to see how meteorological input parameters, time horizons, pre-processing methodology, optimization, and sample size affect the model's complexity and accuracy. Different important parameters & findings of studies are presented in tabular mode. The paper provides key findings based on studied literature to select the optimal model for a specific site. This paper also discusses the metrics used to measure the efficiency of the forecasted model.
The volatile behavior of solar energy is the biggest challenge in its successful integration with existing grid systems. Accurate global horizontal irradiance (GHI) forecasting can resolve this issue and lead to early and effective participation in the energy market. This study proposes a new hybrid deep learning model, namely long short term memory–convolutional neural network (LSTM–CNN), for hourly GHI forecasting, which models the spatio-temporal features by integrating the long short term memory (LSTM) and convolutional neural network (CNN) model. The proposed model is trained with the meteorological data of 23 locations of California State, USA, which includes temperature, precipitation, relative humidity, cloud cover, etc., as input parameters. The proposed hybrid LSTM–CNN model firstly uses LSTM to extract the temporal features from time-series solar irradiance data, followed by CNN, which extracts the spatial features from the correlation matrix of several meteorological variables of target and its neighbor location. The prediction accuracy of the developed model is analyzed rigorously by examining the performance for a year, for four seasons and under three sky conditions. Besides, the proposed LSTM–CNN model shows a forecast skill score in a range of about 37%–45% over few standalone models, including smart persistence, support vector machine, artificial neural network, LSTM, CNN and other hybrid models. The findings of the present work suggest that the proposed hybrid LSTM–CNN model is a reliable alternative for short-term GHI prediction due to its high predictive accuracy under diverse climatic, seasonal and sky conditions.
Data exchange between multiple renewable energy power plant owners can lead to an improvement in forecast skill thanks to the spatio-temporal dependencies in time series data. However, owing to business competitive factors, these different owners might be unwilling to share their data. In order to tackle this privacy issue, this paper formulates a novel privacy-preserving framework that combines data transformation techniques with the alternating direction method of multipliers. This approach allows not only to estimate the model in a distributed fashion but also to protect data privacy, coefficients and covariance matrix. Besides, asynchronous communication between peers is addressed in the model fitting, and two different collaborative schemes are considered: centralized and peer-to-peer. The results for a solar energy dataset show that the proposed method is robust to privacy breaches and communication failures, and delivers a forecast skill comparable to a model without privacy protection.
Intensity of solar irradiance directly affects solar power generation and this makes solar irradiance forecasting a vital process in energy management systems. Existing forecasting systems show positive solar irradiance forecasting performance, but most of them are not accurate in real-life especially when there are fast-moving clouds, causing highly fluctuating solar irradiance profile, which is difficult to predict. Moreover, the requirement to pre-train Artificial Intelligence-based forecasting system has made solar irradiance forecasting impractical as long-hour weather profiles need to be collected prior to deployment. This paper proposes a new artificial intelligent algorithm namely the Regression Enhanced Incremental Self-organising Neural Network (RE-SOINN) for accurate (even for highly fluctuating profile) and adaptive solar irradiance forecasting. This algorithm works by learning the time-series solar irradiance data incrementally and predicting it in real-time. It is novel in terms of enabling the learning of data from discrete (as in the conventional) to continuous using the regression method. The proposed algorithm further improves the prediction accuracy by decomposing the input data into two components (low and high frequency components) before feeding into the RE-SOINNs. Results showed that the proposed algorithm achieves higher accuracy compared to the Persistence model, Exponential Smoothing Model and Artificial Neural Networks.
This paper is an attempt to forecast monthly solar radiation using remote sensing data on a region. Seasonal ARIMA (SARIMA) models are used for simulating and forecasting time series of insolation data from NASA's POWER (Prediction of Worldwide Energy Resources) data archive. Remotely sensed modelled insolation data for 34 years (i.e. January 1984 to December 2017) has been retrieved and analysed for forecasting. Monthly average insolation forecasts of the region around India's capital Delhi have been generated for next four years (i.e. January 2018 to December 2021) and presented in the form of contours obtained using marching square algorithm. The overall accuracy of forecasts in terms of R² (0.9293), Root Mean Square Error (0.3529), Mean Absolute Error (0.2659) and Mean Absolute Percentage Error (6.556) was obtained. The ARIMA model forecasted the maximum insolation values in the months of May (6.52–6.76 KwH/m²/day in year 2018, 6.56–6.8 KwH/m²/day in 2019, 6.6–6.8 KwH/m²/day in 2020 and 6.6–6.84 KwH/m²/day in 2021) and Minimum in the months of January and December (3.2–3.7 KwH/m²/day in January 2018, 3.28–3.52 KwH/m²/day in December 2018). Insolation contours were analysed for identification of potential regions receiving maximum insolation as well as high average annual values of insolation for implementing efficient solar power generation projects. Parts of Haryana and Rajasthan region in study area were found most suitable for such projects.
The prediction of global solar radiation for the regions is of great importance in terms of giving directions of solar energy conversion systems (design, modeling, and operation), selection of proper regions, and even future investment policies of the decision-makers. With this viewpoint, the objective of this paper is to predict daily global solar radiation data of four provinces (Kırklareli, Tokat, Nevşehir and Karaman) which have different solar radiation distribution in Turkey. In the study, four different machine learning algorithms (support vector machine (SVM), artificial neural network (ANN), kernel and nearest-neighbor (k-NN), and deep learning (DL)) are used. In the training of these algorithms, daily minimum and maximum ambient temperature, cloud cover, daily extraterrestrial solar radiation, day length and solar radiation of these provinces are used. The data is supplied from the Turkish State Meteorological Service and cover the last two years (01.01.2018–31.12.2019). To decide on the success of these algorithms, seven different statistical metrics (R², RMSE, rRMSE, MBE, MABE, t-stat, and MAPE) are discussed in the study. The results shows that R², MABE, and RMSE values of all algorithms are ranging from 0.855 to 0.936, from 1.870 to 2.328 MJ/m², from 2.273 to 2.820 MJ/m², respectively. At all cases, k-NN exhibited the worst result in terms of R², RMSE, and MABE metrics. Of all the models, DL was the only model that exceeded the t-critic value. In conclusion, the present paper is reporting that all machine learning algorithms tested in this study can be used in the prediction of daily global solar radiation data with a high accuracy; however, the ANN algorithm is the best fitting algorithm among all algorithms. Then it is followed by DL, SVM and k-NN, respectively.
Solar energy is the most popular resource for power generation among the various available renewable energy alternatives. Solar radiation data are important for solar systems and energy-efficient building designs. Due to the unavailability of measurement, solar radiation prediction models are required. Recently, machine learning techniques were successfully used for predicting solar radiation. However, previous works were mainly focusing on monthly average daily or daily solar radiation. In this study, models for predicting hourly global solar radiation on a horizontal surface were developed based on Multivariate Adaptive Regression Spline (MARS) method. Hourly meteorological data measured in 7 years were used for the study. Sensitivity analysis was conducted using MARS algorithm and the most important variables were selected as inputs of the proposed models. 16 MARS models with different combinations of input variables were proposed. Logistic regression and Artificial Neural Networks (ANN) methods were also used to develop models for comparative study. Finally, the proposed models were evaluated against measurements and compared with existing models. The results showed that the proposed MARS models have good performance in both prediction accuracy and interpretability. The proposed models could be used to estimate effectively the hourly solar radiation according to different combinations of measured variables.
Increasing air pollutants significantly affect the proportion of diffuse (Rd) to global (Rs) solar radiation. This study proposed three new hybrid support vector machines (SVM) with particle swarm optimization algorithm (SVM-PSO), bat algorithm (SVM-BAT) and whale optimization algorithm (SVM-WOA) for predicting daily Rd in air-polluted regions. These models were further compared to standalone SVM, multivariate adaptive regression spline (MARS) and extreme gradient boosting (XGBoost) models. The results showed that models with suspended particulate matter with aerodynamic diameter smaller than 2.5 μm and 10 μm (PM2.5 and PM10) and ozone (O3) produced more accurate daily Rd estimates than those without air pollution parameters, with average relative decreases in root mean square deviation (RMSD) of 11.1%, 10.0% and 10.4% for sunshine duration-based, Rs-based and combined models, respectively. SVM showed better accuracy than XGBoost and MARS. However, compared to SVM, SVM-BAT further enhanced the prediction accuracy and convergence rate in daily Rd modeling, followed by SVM-WOA and SVM-PSO, with relative decreases in RMSD of 2.9%–5.6%, 1.9%–4.9% and 1.1%–3.3%, respectively. The results highlighted the significance of incorporating air pollutants for more accurate estimation of daily Rd in air-polluted regions. Heuristic algorithms, especially BAT, are highly recommended for improving performance of standalone machine learning models.
Many empirical equations and methods have been used and proposed in order to estimate the solar radiation (Rs). In this work, empirical equations, such as Hargreaves method, Artificial Neural Networks (ANN) technology and multi-linear regression methods (MLR) were used to estimate solar radiation. The daily meteorological measurements of air temperature, radiation, humidity and wind velocity from the stations of Aristotle University Farm and Amintaio in Northern Greece were used to derive the solar radiation models. The measurements of Rs were used to derive new and to evaluate existing models. Different combinations of input variables were examined in the ANN models, and in the MLR models different variables were used. The results of RMSE criteria of the examined models showed that they are in the same range with many other models describing Rs as summarized in many review articles. The use of extraterrestrial radiation and the square root of daily difference in temperature in the ANN and MLR models improve the accuracy of the results. The results of ANN models in comparison to MLR models using the same input variables are consistent between them.