Evaluation of the WRF model solar irradiance forecasts in Andalusia (southern Spain)

In this work, we evaluate the reliability of three-days-ahead global horizontal irradiance (GHI) and direct normal irradiance (DNI) forecasts provided by the WRF mesoscale atmospheric model for Andalusia (southern Spain). GHI forecasts were produced directly by the model, while DNI forecasts were obtained based on a physical post-processing procedure using the WRF outputs and satellite retrievals. Hourly time resolution and 3 km spatial resolution estimates were tested against ground measurements collected at four radiometric stations along the years 2007 and 2008. The evaluation was carried out independently for different forecast horizons (1, 2 and 3 days ahead), the different seasons of the year and three different sky conditions: clear, cloudy and overcast. Results showed that the WRF model presents considerable skill in forecasting both GHI and DNI, overall, better than a trivial persistence model. Nevertheless, both MBE and RMSE values presented a marked dependence on the sky conditions and season of the year. Particularly, for 24 h lead time, the MBE of the forecasted GHI was 2% for clear-skies and 18% for cloudy conditions. However, the MBE of the forecasted DNI increased up to about 10% and 75% for clear and cloudy conditions, respectively. Regarding RMSE values, in the case of forecasted GHI, results ranged from below 10% under clear-skies to 50% for cloudy conditions. In the case of forecasted DNI, RMSE ranged from 20% to 100% for clear and cloudy skies, respectively. This proved the higher sensitivity of DNI to the sky conditions. In general, an increment of the MBE and RMSE values with the cloudiness was observed. This reflects a still limited ability of the WRF model to properly forecast cloudy conditions compared to clear skies. Nevertheless, the model was able to accurately forecast steep changes in the sky (cloudiness) conditions. Finally, WRF performed considerable better than the persistence model for clear skies both for GHI and DNI, with relative RMSE values about a half. However, for cloudy conditions, performance was similar.     

Hourly solar irradiance time series forecasting using cloud cover index

We apply time series analysis to forecast next hour solar irradiance including cloud cover effects. Three forecasting methods are proposed using different types of meteorological data as input parameters, namely, global horizontal irradiance (GHI), diffuse horizontal irradiance (DHI), direct normal irradiance (DNI) and cloud cover. The first method directly uses GHI to forecast next hour GHI through additive seasonal decomposition followed by an Auto-Regressive Integrated Moving Average (ARIMA) model. The second method forecasts DHI and DNI separately using additive seasonal decomposition followed by an ARIMA model and then combines the two forecasts to predict GHI using an atmospheric model. The third method considers cloud cover effects. An ARIMA model is used to predict cloud transients. GHI at different zenith angles and under different cloud cover conditions is constructed using nonlinear regression, i.e., we create a look-up table of GHI regression models for different cloud cover conditions. All three methods are tested using data from two weather stations in the USA: Miami and Orlando. It is found that forecasts using cloud cover information can improve the forecast accuracy.     

New challenges in solar energy resource and forecasting in Greece

Aerosols and clouds are the most important constituents in the atmosphere that affect the incoming solar radiation, either directly through absorbing and scattering processes or indirectly by changing the optical properties and lifetime of clouds. Under clear skies, aerosols become the dominant factor that affect the intensity of solar irradiance reaching the ground. Under cloudy skies, the high temporal and spatial variability of cloudiness is the key factor for the estimation of solar irradiance. In this study, recent research activities related to the climatology and the prediction of solar energy in Greece are presented with emphasis on new challenges in the climatology of global horizontal irradiance (GHI) and direct normal irradiance (DNI), the changes of DNI due to the decreasing aerosol optical depth and the short-term (15–240 min) forecasts of solar irradiance with the collaborative use of neural networks and satellite images.     

Day-ahead resource forecasting for concentrated solar power integration

In this work, we validate and enhance previously proposed singe-input direct normal irradiance (DNI) models based on numerical weather prediction (NWP) for intra-week forecasts with over 200,000 hours of ground measurements for 8 locations. Short latency re-forecasting methods to enhance the deterministic forecast accuracies are presented and discussed. The basic forecast is applied to 15 additional locations in North America with satellite-derived DNI data. The basic model outperforms the persistence model at all 23 locations with a skill between 12.4% and 38.2%. The RMSE of the basic forecast is in the range of 204.9 W m⁻² to 309.9 W m⁻². The implementation of stochastic learning re-forecasting methods yields further reduction in error from 204.9 W m⁻² to 176.5 W m⁻². To a great extent, the errors are caused by inaccuracies in the NWP cloud prediction. Improved assessment of atmospheric turbidity has limited impact on reducing forecast errors. Our results suggest that NWP-based DNI forecasts are very capable of reducing power and net-load uncertainty introduced by concentrated solar power plants at all locations in North America. Operating reserves to balance uncertainty in day-ahead schedules can be reduced on average by an estimated 28.6% through the application of the basic forecast.     

Long-term variability of solar direct and global radiation derived from ISCCP data and comparison with reanalysis data

Annual variations of solar radiation at the Earth’s surface may be strong and could seriously harm the return of investment for solar energy projects. This paper analyzes the long-term variability of broadband surface solar radiation based on 18 years of three-hourly satellite observations from the International Satellite Cloud Climatology Project (ISCCP). Direct normal irradiance (DNI) and global horizontal irradiance (GHI) at the surface are derived through radiative transfer calculations, using different physical input parameters describing the actual composition of the atmosphere. Validation of DNI is performed with two years of high resolution Meteosat-derived irradiance. Monthly averages show an average mean bias deviation of −1.7%. Results for DNI from the 18-year time series indicate strong and significant increases for several regions in the subtropics up to +4 W/m² per year, with exception of Australia, where a small decrease in DNI of -1 W/m² per year is observed. Inter-annual variability for DNI is very strong and sometimes exceeds 20%. Comparisons of calculations with and without volcanic aerosol reveal a decrease of up to 16% in annual averages due to volcano eruptions. Changes in GHI are much smaller and less significant. Results show a maximum increase of 0.8 W/m² per year and an annual variability of less than 4%. Volcano eruptions reduce annual averages of GHI by less than 2.2%. The two reanalysis data sets investigated differ strongly from each other and are far off the validated results derived from satellite data. Trends are weaker and less significant or even of opposite sign.     

Temporal variability in direct and global irradiance at various time scales as affected by aerosols

The aerosol optical depth (AOD) is known to be a critical input for radiation modeling purposes, and partially determines the accuracy of modeled direct normal irradiance (DNI) and global horizontal irradiance (GHI). This contribution examines to what extent time variations in AOD also determine the observed variability in DNI, particularly at the daily and longer time scales. Two measures of variability are introduced: the Aerosol Variability Index (AVI) characterizes the magnitude of the variability in AOD over specific periods, from daily to yearly, whereas the Aerosol Sensitivity Index (ASI) relates the magnitude of relative variations in irradiance to absolute variations in AOD. AOD measurements at 180 Aeronet sites over the world are used to obtain clear-sky irradiances with the REST2 radiative model, as well as determinations of ASI and AVI. Large geographic variations exist in AVI, whose largest values are found over western Sahara. The variations of ASI follow a different pattern because it decreases when AOD increases. The variability in GHI is typically 2–4 times lower than that in DNI. On a long-term basis, the normal aerosol-induced variability in DNI is less than ±5% at most sites, but some areas might experience a much larger variability, comparable to that created by large volcanic eruptions. The latest such events predate most current modeled DNI or GHI datasets, making resource assessments potentially too optimistic for bankability if based on such limited data series alone.     

A simple and efficient procedure for increasing the temporal resolution of global horizontal solar irradiance series

The intermittent nature of instantaneous solar radiation has a considerable impact on the nonlinear behavior of solar energy conversion systems. The time resolution of the Numerical Weather Prediction Models (NWPM) or satellite derived solar irradiance data are typically limited to 1-h (at best 15-min). Unfortunately, this resolution is not sufficient in the design and performance of many solar systems. In this study, a new methodology has been developed to increase the temporal resolution of GHI series from 1-h to 1-min. This methodology uses the clearness index k_t (the ratio of GHI to top-of-atmosphere irradiance on the same plane) to characterize the GHI high-frequency dynamics from a 1-year measurement campaign at a given site. The evaluation of the method with 2 years of measured data in different climatic zones has resulted in KSI(%) (Kolmogorov–Smirnov test Integral parameter) and normalized root mean square deviation values below 8.0% and 1.7% respectively for each month, with negligible bias. Indicators of overall performance show an excellent agreement between measured and modeled 1-min GHI data for each month: average values for Nash-Sutcliffe efficiency, Willmott index of agreement and Legates coefficient of efficiency are found to be 0.94, 0.99 and 1.00, respectively.     

Real-time prediction intervals for intra-hour DNI forecasts

We develop a hybrid, real-time solar forecasting computational model to construct prediction intervals (PIs) of one-minute averaged direct normal irradiance for four intra-hour forecasting horizons: five, ten, fifteen, and 20 min. This hybrid model, which integrates sky imaging techniques, support vector machine and artificial neural network sub-models, is developed using one year of co-located, high-quality irradiance and sky image recording in Folsom, California. We validate the proposed model using six-month of measured irradiance and sky image data, and apply it to construct operational PI forecasts in real-time at the same observatory. In the real-time scenario, the hybrid model significantly outperforms the reference persistence model and provides high performance PIs regardless of forecast horizon and weather condition.     

Hybrid intra-hour DNI forecasts with sky image processing enhanced by stochastic learning

We propose novel smart forecasting models for Direct Normal Irradiance (DNI) that combine sky image processing with Artificial Neural Network (ANN) optimization schemes. The forecasting models, which were developed for over 6 months of intra-minute imaging and irradiance measurements, are used to predict 1 min average DNI for specific time horizons of 5 and 10 min. We discuss optimal models for low and high DNI variability seasons. The different methods used to develop these season-specific models consist of sky image processing, deterministic and ANN forecasting models, a genetic algorithm (GA) overseeing model optimization and two alternative methods for training and validation. The validation process is carried over by the Cross Validation Method (CVM) and by a randomized training and validation set method (RTM). The forecast performance for each solar variability season is evaluated, and the models with the best forecasting skill for each season are selected to build a hybrid model that exhibits optimal performance for all seasons. An independent testing set is used to assess the performance of all forecasting models. Performance is assessed in terms of common error statistics (mean bias and root mean square error), but also in terms of forecasting skill over persistence. The hybrid forecast models proposed in this work achieve statistically robust forecasting skills in excess of 20% over persistence for both 5 and 10 min ahead forecasts, respectively.     

Evaluation of decomposition models of various complexity to estimate the direct solar irradiance over Belgium

Solar energy production is directly correlated to the amount of radiation received at a given location. Appropriate information on solar resources is therefore very important for designing and sizing solar energy systems. Concentrated solar power projects and photovoltaic tracking systems rely predominantly on direct normal irradiance (DNI). However, the availability of DNI measurements from surface observation stations has proven to be spatially too sparse to quantify solar resources at most potential sites. Satellite data can be used to calculate estimates of direct solar radiation where ground measurements do not exist. Performance of decomposition models of various complexity have been evaluated against one year of in situ observations recorded on the roof of the radiometric tower of the Royal Meteorological Institute of Belgium in Uccle, Brussels. Models were first evaluated on a hourly and sub-hourly basis using measurements of global horizontal irradiance (GHI) as input. Second, the best performing ground-based decomposition models were used to extract the direct component of the global radiation retrieved from Meteosat Second Generation (MSG) images. Results were then compared to direct beam estimations provided by satellite-based diffuse fraction models and evaluated against direct solar radiation data measured at Uccle. Our analysis indicates that valuable DNI estimation can be derived from MSG images over Belgium regardless of the satellite retrieved GHI accuracy. Moreover, the DNI retrieval from MSG data can be implemented on an operational basis.     

Using direct normal irradiance models and utility electrical loading to assess benefit of a concentrating solar power plant

The objective of this paper was to determine if three different direct normal irradiance (DNI) models were sufficiently accurate to determine if concentrating solar power (CSP) plants could meet the utility electrical load. DNI data were measured at three different laboratories in the United States and compared with DNI calculated by three DNI models. In addition, utility electrical loading data were obtained for all three locations. The DNI models evaluated were: the Direct Insolation Simulation Code (DISC), DIRINT, and DIRINDEX. On an annual solar insolation (e.g. kW h/m²) basis, the accuracy of the DNI models at all three locations was within: 7% (DISC), 5% (DIRINT), and 3% (DIRINDEX). During the three highest electrical loading months at the three locations, the monthly accuracy varied from: 0% to 16% (DISC), 0% to 9% (DIRINT), and 0% to 8% (DIRINDEX). At one location different pyranometers were used to measure GHI, and the most expensive pyranometers did not improve the DNI model monthly accuracy. In lieu of actually measuring DNI, using the DIRINT model was felt to be sufficient for assessing whether to build a CSP plant at one location, but use of either the DIRINT or DIRINDEX models was felt to be marginal for the other two locations due to errors in modeling DNI for utility peak electrical loading days – especially for partly cloudy days.     

Intra-hour forecasting with a total sky imager at the UC San Diego solar energy testbed

A method for intra-hour, sub-kilometer cloud forecasting and irradiance nowcasting using a ground-based sky imager at the University of California, San Diego is presented. Sky images taken every 30 s were processed to determine sky cover using a clear sky library and sunshine parameter. From a two-dimensional cloud map generated from coordinate-transformed sky cover, cloud shadows at the surface were estimated. Limited validation on four partly cloudy days showed that (binary) cloud conditions were correctly nowcast 70% of the time for a network of six pyranometer ground stations spread out over an area of 2 km². Cloud motion vectors were generated by cross-correlating two consecutive sky images. Cloud locations up to 5 min ahead were forecasted by advection of the two-dimensional cloud map. Cloud forecast error increased with increasing forecast horizon due to high cloud cover variability over the coastal site.     

Evaluation of numerical weather prediction for intra-day solar forecasting in the continental United States

Numerical weather prediction (NWP) is generally the most accurate tool for forecasting solar irradiation several hours in advance. This study validates the North American Model (NAM), Global Forecast System (GFS), and European Centre for Medium-Range Weather Forecasts (ECMWF) global horizontal irradiance (GHI) forecasts for the continental United States (CONUS) using SURFRAD ground measurement data. Persistence and clear sky forecasts are also evaluated. For measured clear conditions all NWP models are biased by less than 50 W m⁻². For measured cloudy conditions these biases can exceed 200 W m⁻² near solar noon. In general, the NWP models (especially GFS and NAM) are biased towards forecasting clear conditions resulting in large, positive biases.Mean bias errors (MBE) are obtained for each NWP model as a function of solar zenith angle and forecast clear sky index, kt^∗, to derive a bias correction function through model output statistics (MOS). For forecast clear sky conditions, the NAM and GFS are found to be positively biased by up to 150 W m⁻², while ECMWF MBE is small. The GFS and NAM forecasts were found to exceed clear sky irradiances by up to 40%, indicating an inaccurate clear sky model. For forecast cloudy conditions (kt^∗ < 0.4) the NAM and GFS models have a negative bias of up to −150 W m⁻². ECMWF forecasts are most biased for moderate cloudy conditions (0.4 < kt^∗ < 0.9) with an average over-prediction of 100 W m⁻².MOS-corrected NWP forecasts based on solar zenith angle and kt^∗ provide an important baseline accuracy to evaluate other forecasting techniques. MOS minimizes MBE for all NWP models. Root mean square errors for hourly-averaged daytime irradiances are also reduced by 50 W m⁻², especially for intermediate clear sky indices. The MOS-corrected GFS provides the best solar forecasts for the CONUS with an RMSE of about 85 W m⁻², followed by ECMWF and NAM. ECMWF is the most accurate forecast in cloudy conditions, while GFS has the best clear sky accuracy.     

Forecasting of global and direct solar irradiance using stochastic learning methods, ground experiments and the NWS database

We develop and validate a medium-term solar irradiance forecasting model by adopting predicted meteorological variables from the US National Weather Service’s (NWS) forecasting database as inputs to an Artificial Neural Network (ANN) model. Since the inputs involved are the same as the ones available from a recently validated forecasting model, we include mean bias error (MBE), root mean square error (RMSE), and correlation coefficient (R²) comparisons between the more established forecasting model and the proposed ones. An important component of our study is the development of a set of criteria for selecting relevant inputs. The input variables are selected using a version of the Gamma test combined with a genetic algorithm. The solar geotemporal variables are found to be critically important, while the most relevant meteorological variables include sky cover, probability of precipitation, and maximum and minimum temperatures. Using the relevant input sets identified by the Gamma test, the developed forecasting models improve RMSEs for GHI by 10-15% over the reference model. Prediction intervals based on regression of the squared residuals on the input variables are also derived.     

MEO shadowring method for measuring diffuse solar irradiance: Corrections based on sky cover

The present paper deals with atmospheric corrections factors proposed as a function of the atmospheric transmissivity in order to correct the diffuse solar irradiance measured with the Melo-Escobedo-Oliveira Shadowring Measuring Method (MEO shadowring Method). Global irradiance was measured by an Eppley-PSP pyranometer; direct normal irradiance by an Eppley-NIP pyrheliometer fitted to a ST-3 sun tracking device and diffuse irradiance by an Eppley-PSP pyranometer fitted to a MEO shadowring. The Solar Radiometric Laboratory at Sao Paulo State University provided the measurements during the years 1996–2005. Two correction models for diffuse solar irradiance were proposed: All Sky Correction Model (ASC Model) and Sky Cover Correction Model (SCC Model). The MBE and RMSE statistical indicators performed the validations. The correction models showed results in the same order of magnitude: ASC Model showed 0.81% deviation, while SCC Model showed 0.66% deviation. Therefore, the correction models proposed as a function of the sky covering (atmospheric transmissivity) were efficient to correct the isotropic diffuse irradiance, approaching the measured and reference diffuse irradiance less than 1%. Corrections show dependence on sky coverage and seasonality. The results presented that the sky cover corrections improve the MEO shadowring method, allowing the generation of a reliable global, direct and diffuse radiation database without high financial investments.     

A simple approach to the synthetic generation of solar irradiance time series with high temporal resolution

Modeling the performance of some concentrating solar systems for thermal power plants may require high temporal resolution irradiance as input, in order to account for the impact of the cloud transient effects. This work proposes a simple method of generating synthetic irradiance of 10-min intervals from the hourly mean values. Boundary conditions are imposed to preserve the expected behavior under clear sky situations. The procedure consists basically on adding a random fluctuation, which characteristic amplitude depends on the sky conditions, to the hourly interpolated values. The assessment of the method with ground data have shown to main aspects to remark: daily and monthly means from the synthetic data are below 5% of root mean squared deviation compared to the original time series; despite the noticeable uncertainty in the 10-min synthetic irradiance values, the dynamic behavior of the fluctuations is comparable to the original data.     

A case study to prepare for the utilization of aerosol forecasts in solar energy industries

Precise aerosol information is indispensable in providing accurate clear sky irradiance forecasts, which is a very important aspect in solar facility management as well as in solar and conventional power load prediction. In order to demonstrate the need of detailed aerosol information, direct irradiance derived from Aerosol Robotic Network (AERONET) ground based measurements of aerosol optical depth (AOD) was compared in a case study over Europe to irradiance calculated using a standard aerosol scenario. The analysis shows an underestimation of measurement-derived direct irradiance by the scenario-derived direct irradiance for locations in Northern Europe and an overestimation for the Mediterranean region.Forecasted AOD of the European Dispersion and Deposition Model (EURAD) system was validated against ground based AERONET clear sky AOD measurements for the same test period of February 15th to 22nd, 2004. For the time period analyzed, the modelled AOD forecasts of the EURAD system slightly underestimate ground based AERONET measurements. To quantify the effects of varying AOD forecast quality in their impact on the application in solar energy industry, measured and forecasted AOD were used to calculate and compare direct, diffuse, and global irradiance. All other influencing variables (mainly clouds and water vapour) are assumed to be modelled and measured correctly for this analysis which is dedicated to the specific error introduced by aerosol forecasting. The underestimated AOD results in a mean overestimation of direct irradiance of +28 W/m² (+12%), whereas diffuse irradiance is generally underestimated (−19 W/m² or −14%). Mean global irradiance values where direct and diffuse irradiance errors compensate each other are very well represented (on average +9 W/m² or +2%).     

A new approach to estimating the diffuse irradiance on inclined surfaces

The diffuse irradiance on an inclined surface is usually estimated from the hourly horizontal irradiance measurements with a slope irradiance model. It is also possible to calculate the slope irradiance by integrating the sky radiance distribution generated with a sky radiance model. In this paper, five slope irradiance models and six sky distribution models are compared with the hourly irradiance measurements on 24 inclined surfaces in Turku, Finland (60°27′N, 22°18′E). Of the sky distribution models, the Perez all-weather sky model agrees best with the measurements. Of the slope irradiance models, the Reindl model has the lowest average mean bias difference (MBD), but the Perez slope irradiance model gives the lowest root mean square difference (RMSD) for all but one of the 24 surface orientations. The average RMSD for the Perez all-weather sky model is 1.5 percentage points lower than for the Perez slope irradiance model.     

基于改进CNN的光热电场太阳直接法向辐射预测研究

为了在实际运行中更好地利用光热电站,文章建立了一种基于改进卷积神经网络的光热电场太阳直接法向辐射的预测模型。首先,通过分析光热发电系统的运行机理,得到影响光热发电系统出力的主要因素是太阳直接法向辐射,并根据太阳直接法向辐射特点选用卷积神经网络对其进行预测;然后,针对卷积神经网络在实际应用过程中存在的预测精度较低和训练时间较长的问题,引入带有稀疏约束的损失函数和自适应学习率思想,并提出一种改进卷积神经网络;最后,利用改进卷积神经网络建立了光热电场太阳直接法向辐射的预测模型。模拟结果表明:文章提出的改进卷积神经网络能够解决一般卷积神经网络在实际应用中存在的预测精度较低和训练速度较慢的问题;基于改进卷积神经网络的预测模型可以较准确地预测出太阳直接法向辐射的变化趋势及其数值。     

Practical models to estimate horizontal irradiance in clear sky conditions: Preliminary results

The Argentinean Northwest (ANW) is a high altitude region located alongside Los Andes Mountains. The ANW is also one of the most insolated regions in the world due to its altitude and particular climate. However, the characterization of the solar resource in the region is incomplete as there are no stations to measure solar radiation continuously and methodically. With irradiance data recently having been measured at three sites in the Salta Province, a study was carried out that resulted in a practical model to quickly and efficiently estimate the horizontal irradiance in high altitude sites in clear sky conditions. This model uses the altitude above sea level (A) as a variable and generates a representative clearness index as a result (k_t-R) that is calculated for each site studied. This index k_t-R is then used with the relative optical air mass and the extraterrestrial irradiance to estimate the instantaneous clearness index (k_t). Subsequently, the index k_t-R is corrected by introducing the atmospheric pressure in the definition of relative optical air mass proposed by Kasten. The results are satisfactory as errors in the irradiance estimations with respect to measured values do not exceed 5% for pressure corrected air masses AM_c < 2. This model will be used in a feasibility study to locate sites for the installation of solar thermal power plants in the ANW. A prototype of a CLFR solar power plant is being built in the INENCO Campus, at the National University of Salta.