Fig 4 - available via license: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
Content may be subject to copyright.
An example of cloud motion vectors derived from (a) cloud optical thickness, (b) cloud top height, (c) effective radius, and (d) the combined CMVs on 16 June 2016 at 12 UTC. The color of the vectors (blue, light blue, green) indicates the time steps at which the CMV were derived. The length of the vector is proportional to the wind speed. The circles show the end positions of the last vectors. In (d) the vectors are sampled at about 50 km × 50 km resolution for plotting. The area roughly spans 0-9°E and 48-54°N. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Source publication
A surface solar radiation forecast algorithm is developed using cloud physical properties from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on board of the Meteosat Second Generation (MSG) geostationary satellite. The novelty of the algorithm is the derivation of cloud motion vectors using cloud physical properties. The solar radiatio...
Contexts in source publication
Context 1
... example of the atmospheric motion vectors derived from different cloud properties is shown in Fig. 4. The area is about 0-9°E, 48-54°N. The end positions of the last vectors in 5 consecutive images are indicated with white circles. The vectors derived from different time steps are in different colors (from blue to green), which correspond to the CMVs of V 1 , V 2 , V 3 in Fig. 2. The length of the vector or arrow is proportional to ...
Context 2
... more motion vectors are derived from CTH than from the other cloud properties, which suggests that CTH has more fine structures than COT or R eff . The vectors derived from different cloud properties have almost no overlap in locations. Normally, between 30 and 180 cloud motion vectors can be derived from five satellite images. As shown in Fig. 4(d) the final CMV field is quite smooth and ...
Citations
... Their results show that the WRF-Solar model was able to accurately forecast GHI under different atmospheric conditions. On the other hand, statistical approaches such as ML models were more reliable in forecasting GHI than NWP models for lead times up to 6 h [72]. In Africa, most GHI forecasts are based on the ML models. ...
Accurate global horizontal irradiance (GHI) forecasting is critical for integrating solar energy into the power grid and operating solar power plants. The Weather Research and Forecasting model with its solar radiation extension (WRF-Solar) has been used to forecast solar irradiance in different regions around the world. However, the application of the WRF-Solar model to the prediction of GHI in West Africa, particularly Ghana, has not yet been investigated. The aim of this study is to evaluate the performance of the WRF-Solar model for predicting GHI in Ghana, focusing on three automatic weather stations (Akwatia, Kumasi and Kologo) for the year 2021. We used two one-way nested domains (D1 = 15 km and D2 = 3 km) to investigate the ability of the fully coupled WRF-Solar model to forecast GHI up to 72-hour ahead under different atmospheric conditions. The initial and lateral boundary conditions were taken from the ECMWF high-resolution operational forecasts. Our findings reveal that the WRF-Solar model performs better under clear skies than cloudy skies. Under clear skies, Kologo performed best in predicting 72-hour GHI, with a first day nRMSE of 9.62 %. However, forecasting GHI under cloudy skies at all three sites had significant uncertainties. Additionally, WRF-Solar model is able to reproduce the observed GHI diurnal cycle under high AOD conditions in most of the selected days. This study enhances the understanding of the WRF-Solar model’s capabilities and limitations for GHI forecasting in West Africa, particularly in Ghana. The findings provide valuable information for stakeholders involved in solar energy generation and grid integration towards optimized management in the region.
... CMVs are computed for the features to estimate their future paths. Satellite CMVbased methods have been shown to outperform numerical weather prediction methods for forecasting cloud motion for lead times of up to 4 h [24]. Starting from [25] by using a cross-correlation method to detect movements, many works in the literature exploited satellite maps to derive cloud motion vectors. ...
... In [28], different operational CMV methods are compared to forecast clear-sky index maps, and the best performing method is selected for each pixel. More recently, advection-based approaches have been used to forecast cloud properties such as cloud height and thickness [29,24], while in [22], the authors computed the advection field directly on the SEVIRI different spectral channel maps and then converted them into solar radiation with a radiative transfer model. However, the most common features used to extract the advection fields are the cloudiness indices [30]. ...
... This is surprising because forecast methods using satellite data and CMVs hold great potential for higher accuracy intraday forecasts with well-characterized uncertainties [32]. Several satellite-CMV-based SSR forecasting methods have been demonstrated in the past decade (e.g., [33,34,30,24]) but they provide deterministic forecasts that lack the capability to characterize forecast uncertainties. ...
Solar energy supply is usually highly volatile, limiting its integration into the power grid. Accurate probabilistic intraday forecasts of solar resources are essential to increase the share of photovoltaic (PV) energy in the grid and enable cost-efficient balancing of power demand and supply. Solar PV production mainly depends on down-welling surface solar radiation (SSR). By estimating SSR from geostationary satellites, we can cover large areas with high spatial and temporal resolutions, allowing us to track cloud motion. State-of-the-art probabilistic forecasts of solar resources from satellite imagery account only for the advective motion of clouds. They do not consider the evolution of clouds over time, their growth, and dissipation, even though these are major sources of forecast uncertainty. To address the uncertainty of cloudiness evolution, we present SolarSTEPS, the first optical-flow probabilistic model able to simulate the temporal variability of cloudiness. We demonstrate that forecasting
the autocorrelated scale-dependent evolution of cloudiness outperforms state-of-the-art probabilistic advection-based forecasts by 9% in continuous ranked probability score (CRPS). This corresponds to an extension of the forecast lead time by about 45 min at constant CRPS. Our work is motivated by the scale-dependent predictability of cloud growth and decay. We demonstrate that cloudiness is more variable in time at smaller spatial scales than at larger ones. Specifically, we show that the temporal autocorrelation of cloudiness is related to its spatial scale by an inverse power law. We also demonstrate that decomposing cloudiness into multiple spatial scales in the forecasts further improves the forecast skill, reducing the CRPS by 10% and the RMSE by 9%.
... The approach uses ensemble empirical mode decomposition (EEMD) with k-means clustering, and parameters are optimized using gravitational search algorithm (GSA). Wang et al. (2019) forecast global horizontal irradiance (or GHI) and direct normal irradiance (DNI) by utilizing satellite imagery of movement of clouds and by tracing the corresponding vectors of their movement. Basically, the researchers used the numerical weather prediction approach for prediction of solar irradiance. ...
Renewable energy plays an important role in the power mix of India being sustainable and environmental source of energy. In this study, modified fuzzy Q-learning (MFQL)-based solar radiation forecasting has been proposed to forecast 30-min-ahead solar irradiance. Application of MFQL is novel in this field, as it uses reinforcement learning and model-free environment. Raw data have been collected for four Indian cities in the state of Rajasthan, i.e. Jodhpur, Ajmer, Jaipur and Kota via the data portal of National Institute of Wind Energy and Wind Resource (NIWE). Empirical mode decomposition (EMD) has been used as the data pre-processing technique, and relevant features are extracted from Pearson’s correlation coefficient. The results obtained from the MFQL forecaster are promising with forecasting accuracy of 92.38% for winter, 93.73% for summer, 91.54% for monsoon and 92.05% for autumn season for the city of Ajmer, and similar results have been obtained for other cities as well. MFQL lends itself as an effective tool for forecasting of seasonal solar irradiance. Proposed prediction model can be effectively utilized for solar irradiance forecasting and for optimal generation of power from incident solar radiation.
... CMVs are computed for the features to estimate their future paths. Satellite CMV-based methods have been shown to outperform numerical weather prediction methods for forecasting cloud motion for lead times of up to 4 hours (Wang et al., 2019). Starting from Leese et al., 1971 by using a cross-correlation method to detect movements, many works in the literature exploited satellite maps to derive cloud motion vectors. ...
... In Perez et al., 2010, different operational CMV methods are compared to forecast clear-sky index maps, and the best performing method is selected for each pixel. More recently, similar approaches have been used to forecast cloud properties such as cloud height and thickness (Batlles et al., 2014, Wang et al., 2019, while in Gallucci et al., 2018, the authors computed the advection field directly on the SEVIRI different spectral channel maps and then converted them into solar radiation with a radiative transfer model. However, the most common features used to extract the advection fields are the cloudiness indices (Urbich et al., 2018). ...
... This is surprising because forecast methods using satellite data and CMVs hold great potential for higher accuracy intraday forecasts with wellcharacterized uncertainties (Yang et al., 2018). Several satellite-CMV-based SSR forecasting methods have been demonstrated in the past decade (e.g., Wang et al. 2019;Urbich et al., 2018Coimbra et al., 2013) but they provide deterministic forecasts that lack the capability to characterize forecast uncertainties. Probabilistic methods for CMV-based SSR forecasting have barely been developed up to now. ...
Solar energy supply is usually highly volatile which limits its integration in the power grid. Accurate probabilistic intraday forecasts of solar resources are essential to increase the share of photovoltaic (PV) energy in the grid and enable cost-efficient balancing of power demand and supply. Solar PV production mainly depends on downwelling surface solar radiation (SSR). By estimating SSR from geostationary satellites, we can cover large areas with high spatial and temporal resolutions, allowing us to track cloud motion. State-of-the-art probabilistic forecasts of solar resources from satellite imagery account only for the advective motion of clouds. They do not consider the evolution of clouds over time, their growth, and dissipation, even though these are major sources of forecast uncertainty. To address the uncertainty of cloudiness evolution, we present SolarSTEPS, the first optical-flow probabilistic model able to simulate the temporal variability of cloudiness. We demonstrate that forecasting the autocorrelated scale-dependent evolution of cloudiness outperforms state-of-the-art probabilistic advection-based forecasts by 9% in continuous ranked probability score (CRPS). This corresponds to an extension of the forecast lead time by about 45 minutes at constant CRPS. Our work is motivated by the scale-dependent predictability of cloud growth and decay. We demonstrate that cloudiness is more variable in time at smaller spatial scales than at larger ones. Specifically, the temporal autocorrelation of cloudiness is related to its spatial scale by a rational function. We also demonstrate that decomposing cloudiness into multiple spatial scales in the forecasts further improves the forecast skill, reducing the CRPS by 10% and the RMSE by 9%.
... Intra-hour GHI forecast employing a cloud retrieval technique to develop a physics based smart persistence model is improved in [9], and an algorithm using cloud physica properties for intra-day GHI and DNI forecasting with time horizons of 0-4 h at a 15 mi temporal resolution is developed in [10]. A GHI forecasting model based on satellite dat from Finland with a forecast horizon of 4 h and a 15 min temporary resolution is devel oped and validated in [11], while the error obtained from the Japan meteorological agenc mesoscale model in the hourly-averaged GHI forecasts from 2008 to 2012 is assessed in [12]. ...
... Intra-hour GHI forecast employing a cloud retrieval technique to develop a physicsbased smart persistence model is improved in [9], and an algorithm using cloud physical properties for intra-day GHI and DNI forecasting with time horizons of 0-4 h at a 15 min temporal resolution is developed in [10]. A GHI forecasting model based on satellite data from Finland with a forecast horizon of 4 h and a 15 min temporary resolution is developed and validated in [11], while the error obtained from the Japan meteorological agency mesoscale model in the hourly-averaged GHI forecasts from 2008 to 2012 is assessed in [12]. ...
... Short-term irradiance forecasting [9] Intra-hour GHI cloud retrieval technique to develop a physics-based smart persistence model [10] Intra-day GHI and DNI algorithm using cloud physical properties [11] A 15 min GHI forecasting model [12] Hourly-averaged GHI forecasts [13] Hourly GHI and DNI clear-sky irradiance vs. RRTMG physical radiative transfer model [14] Hourly and daily GHI from mesoscale atmospheric weather research forecasting model [15] Hourly GHI with a three-dimensional meteorology-chemistry model including a treatment of aerosols [16] Hourly GHI exponential smoothing model with decomposition methods [17] [25,26] Hourly GHI ANN models [27] Mean daily GHI with ANN models [28] A 500 ms-5 min GHI based on k-means algorithm [29] A 5-30 min GHI and DNI based on the k-nearest neighbours algorithm [30] A 30 min-5 h GHI Gaussian process regression method [31] Daily GHI with ANN models for for 25 Moroccan cities [32] Daily GHI with empirical and machine learning models for 5 Moroccan cities [33] Monthly mean daily GHI using time series models [34] Daily GHI with hybrid ARIMA-ANN model for 3 cities in Morocco [35] Daily GHI with ANN models for 35 Moroccan, Algerian, Spanish and Mauritian cities ...
The prediction and characterization of solar irradiation relies mostly on either the use of complex models or on complicated mathematical techniques, such as artificial neural network (ANN)-based algorithms. This mathematical complexity might hamper their use by businesses and project developers when assessing the solar resource. In this study, a simple but comprehensive methodology for characterizing the solar resource for a project is presented. It is based on the determination of the best probability distribution function (PDF) of the solar irradiation for a specific location, assuming that the knowledge of statistical techniques may be more widely extended than other more complex mathematical methods. The presented methodology was tested on 23 cities across Morocco, given the high interest in solar investments in the country. As a result, a new database for solar irradiation values depending on historical data is provided for Morocco. The results show the great existing variety of PDFs for the solar irradiation data at the different months and cities, which demonstrates the need for undertaking a proper characterization of the irradiation when the assessment of solar energy projects is involved. When it is simply needed to embed the radiation uncertainty in the analysis, as is the case of the techno-economic valuation of solar energy assets, the presented methodology can reach this objective with much less complexity and less demanding input data. Moreover, its application is not limited to solar resource assessment, but can also be easily used in other fields, such as meteorology and climate change studies.
... These changes cause significant fluctuations in the overall power output from the solar PV plant due to variations in the spectral content of light and the angle from which the light is incident on the earth's surface [27]. Local variations specific to location are key contributors to power fluctuations and they include the effect of clouds, seasons, and the length of the daytime [28]. Desert regions have low local atmospheric variations while equatorial regions exhibit low variability between seasons [29]. ...
The increase in penetration of solar photovoltaics into the traditional grid and the accelerating growth of smart grids have introduced new challenges to grid stability. Forecasting the output power from solar PV systems and time-based analysis for the performance characteristics of solar PV under different weather conditions is essential to improve the grid stability. The generated PV power is intermittent in nature and is influenced by meteorological parameters such as pressure, temperature, relative humidity, and solar zenith angle. With the influence of the above parameters, a novel power forecasting model has been developed using Supervised Machine Learning Algorithm. The historical weather data of a given location have been fetched from National Solar Radiation Database (NSRDB) with the corresponding location coordinates. Multivariate data are used as inputs to train a Decision Tree Regression Model in order to predict the solar irradiance parameters such as Global Horizontal Irradiance, Direct Normal Irradiance, and Diffuse Horizontal Irradiance which are essential to calculate the output power harnessed from the grid-connected PV system. The results are favorable for the application and have depicted minimal deviation with an average accuracy of 86.02%. This technique also rules out the need of hardware power prediction modules, favoring a cost-efficient methodology.
... For forecast horizons ranging from a few minutes to 6 h, statistical models with on-site solar irradiance observations as input are more suitable [10,11]. Additionally, cloud motion, provided by satellite images analysis, can be extrapolated to the upcoming few hours and allows to obtain good forecasts for forecast horizons up to 6 h [12,13]. However, techniques based on sky imagers or satellite images are very complex and involve many steps [14]. ...
With the development of predictive management strategies for power distribution grids, reliable information on the expected photovoltaic power generation, which can be derived from forecasts of global horizontal irradiance (GHI), is needed. In recent years, machine learning techniques for GHI forecasting have proved to be superior to classical approaches. This work addresses the topic of multi-horizon forecasting of GHI using Gaussian process regression (GPR) and proposes an in-depth study on some open questions: should time or past GHI observations be chosen as input? What are the appropriate kernels in each case? Should the model be multi-horizon or horizon-specific? A comparison between time-based GPR models and observation-based GPR models is first made, along with a discussion on the best kernel to be chosen; a comparison between horizon-specific GPR models and multi-horizon GPR models is then conducted. The forecasting results obtained are also compared to those of the scaled persistence model. Four performance criteria and five forecast horizons (10 min, 1 h, 3 h, 5 h, and 24 h) are considered to thoroughly assess the forecasting results. It is observed that, when seeking multi-horizon models, using a quasiperiodic kernel and time as input is favored, while the best horizon-specific model uses an automatic relevance determination rational quadratic kernel and past GHI observations as input. Ultimately, the choice depends on the complexity and computational constraints of the application at hand.
... For example, current state-of-the-art research for predicting solar radiation in the Netherlands has a relative root mean square error of 50% for a 3 hours forecast. Only on sunny days this forecast is better [19]. A DSM algorithm should therefore be able to handle large uncertainties in the prediction of solar power. ...
... SSI nowcasting can be achieved using a combination of CMV information and SSI estimates. Numerous SSI nowcasting studies have been conducted based on satellite sensors (e.g., Spinning Enhanced Visible and Infrared Imager (SEVIRI)) (Arbizu- Barrena et al., 2017;Gallucci et al., 2018;Mouhamet et al., 2018;Wang et al., 2019). Fengyun-4A (FY-4A), launched by China in 2016, is a new generation geostationary meteorological satellite. ...
Surface solar irradiance (SSI) nowcasting (0–3 h) is an effective way to overcome the intermittency of solar energy and to ensure the safe operation of grid-connected solar power plants. In this study, an SSI estimate and nowcasting system was established using the near-infrared channel of Fengyun-4A (FY-4A) geostationary satellite. The system is composed of two key components: The first is a hybrid SSI estimation method combining a physical clear-sky model and an empirical cloudy-sky model. The second component is the SSI nowcasting model, the core of which is the derivation of the cloud motion vector (CMV) using the block-matching method. The goal of simultaneous estimation and nowcasting of global horizontal irradiance (GHI) and direct normal irradiance (DNI) is fulfilled. The system was evaluated under different sky conditions using SSI measurements at Xianghe, a radiation station in the North China Plain. The results show that the accuracy of GHI estimation is higher than that of DNI estimation, with a normalized root-mean-square error (nRMSE) of 22.4% relative to 45.4%. The nRMSE of forecasting GHI and DNI at 30–180 min ahead varied within 25.1%–30.8% and 48.1%–53.4%, respectively. The discrepancy of SSI estimation depends on cloud occurrence frequency and shows a seasonal pattern, being lower in spring—summer and higher in autumn—winter. The FY-4A has great potential in supporting SSI nowcasting, which promotes the development of photovoltaic energy and the reduction of carbon emissions in China. The system can be improved further if calibration of the empirical method is improved.
... Solar radiation is the most important factor to affect PV output; hence, its forecast has been extensively investigated [6][7][8]. However, the forecast skill of various methods always depends on the local weather and the accuracy under partly cloudy conditions is hardly satisfactory, with a relative root mean square error (rRMSE) of 30%-70% at hourly scale [9][10][11]. In this context, enhancing the forecast skill for solar radiation has ongoing motivation and is an open issue of concern. ...
... More examples can be found in review literatures [4,14]. However, the learning process of machine learning based methods is usually disrupted by the stochastic fluctuations of solar radiation associated with the variations of cloud cover and other factors [11,20]. Long short-term memory (LSTM) that remembers historical information for subsequent forecast alleviates such disruption to a certain extent [17,21,22]. ...
... It has been proved that the integration of satellite-derived cloud parameters improves the forecast skill for hourly solar radiation [27,28]. The prevailing principle is to infer the speed and trajectory of cloud motion through a similarity analysis of two or more consecutive images [11,29]. Assuming cloud motions remain the same for the next few hours, the cloud positions can be determined, and the subsequent solar radiation is estimated based on the physical mechanism of the interactions between clouds and solar radiation [26]. ...
Accurate output forecasts are essential for photovoltaic projects to achieve stable power supply. Traditional forecasts based on ground observation time series are widely troubled by the phase lag issue due to the incomplete consideration of the impacts of cloud motion. With the consensus that this issue can be addressed by introducing satellite-derived cloud information, we propose an innovative framework that integrates ground and satellite observations through deep learning to enhance PV output forecasts. Cloud motion patterns are captured from satellite observations using convolutional neural networks, and the long-range spatio-temporal cloud impacts on subsequent PV outputs are established by long short-term memory network. The forecast accuracy of real-time PV output is significantly improved, with a minimum (maximum) relative root mean square error of 16% (29%). The ratio of phase lag is reduced to 15% on average. This work provides a potential for alleviating the power intermittency of solar PV system and making advance planning in solar energy utilization.
