Irradiance Variability Research Articles

ECLIPSE: Envisioning CLoud Induced Perturbations in Solar Energy

Efficient integration of solar energy into the electricity mix depends on a reliable anticipation of its intermittency. A promising approach to forecasting the temporal variability of solar irradiance resulting from the cloud cover dynamics is based on the analysis of sequences of ground-taken sky images or satellite observations. Despite encouraging results, a recurrent limitation of existing deep learning approaches lies in the ubiquitous tendency of reacting to past observations rather than actively anticipating future events. This leads to a frequent temporal lag and limited ability to predict sudden events. To address this challenge, we introduce ECLIPSE, a spatio-temporal neural network architecture that models cloud motion from sky images to not only predict future irradiance levels and associated uncertainties, but also segmented images, which provide richer information on the local irradiance map. We show that ECLIPSE anticipates critical events and reduces temporal delay while generating visually realistic futures. The model characteristics and properties are investigated with an ablation study and a comparative study on the benefits and different ways to integrate auxiliary data into the modelling. The model predictions are also interpreted through an analysis of the principal spatio-temporal components learned during network training.

Fragility of marine photosynthesis

Ecosystem fragility is an often used term in oceanography yet to this day it lacks a precise and widely accepted definition. Defining and subsequently quantifying fragility would be of great value, for such measures could be used to objectively ascertain the level of risk marine ecosystems face. Risk assessments could further be used to define the level of protection a given ocean region requires from economic activity, such as fisheries. With this aim we introduce to the oceanographic literature the concepts of marginal production and fragility, which we define for marine photosynthesis, the base of the oceanic food web. We demonstrate that marine photosynthesis is always fragile with respect to light, implying variability in surface irradiance acts unfavourably on biomass. We also demonstrate that marine photosynthesis can be both fragile and antifragile with respect to the mixed-layer depth, implying variability in mixed-layer depth can act both favourably and unfavourably on biomass. Quantification of marginal production and fragility is presented on data from two open ocean stations: Hawaii Ocean Time Series and Bermuda Atlantic Time-Series Study. Seasonal cycle of biomass is modelled and the effects of primary production fragility are analysed. A new tipping point for marine phytoplankton is identified in the form of a depth horizon. Using the new definitions presented here a rich archive of data can be used straightforwardly to quantify primary production fragility. The definitions can also be used to predict when primary production enters the fragile state during the seasonal cycle.

Sun-blocking index from sky image to estimate solar irradiance

To optimize the balance between indoor comfort and minimal energy usage, several automated methods have been proposed for smart window control systems. However, owing to the high variability of solar irradiance, it is difficult for window controllers to operate properly. The variability of solar irradiance is mostly caused by cloud motion in the sky. Cloud information can be obtained from human observations, satellite images, and sky images, and parameters such as cloud amount, cloud fraction, and cloud cover (CC) have been conventionally used for considering cloud effects. These conventional parameters focus only on the amount of cloud over the entire sky while failing to account for the amount of cloud locally around the sun disk. Consequently, rapid changes in the cloud at short time scales have not been considered properly, and the high variability of solar irradiance has not been addressed accurately. Therefore, in this study, a new parameter, the sun-blocking index (SBI), is introduced to model the partial blocking of the sun by clouds and hence enhance the estimation accuracy of solar irradiance under fickle weather. In addition, an image processing model to calculate SBI using sky images was proposed. Once the correlation between the clearness index and SBI is established, the SBI-dependent clearness index is applied to solar radiation models. The results showed that the global horizontal irradiance was estimated with a relative root-mean-square error (rRMSE) of 16.21%. Remarkably, when direct normal irradiance (DNI) was estimated, the proposed model outperformed conventional models with rRMSE of less than 28%. • Partial blocking of the sun by cloud is considered for assessing solar irradiance. • The sun-blocking index (SBI) is proposed to measure the partial blocking. • A methodology for calculating the SBI from sky images is developed. • The relationship between SBI and clearness index is investigated. • Solar irradiance is more accurately estimated via the SBI-dependent clearness index.

Evaluation of artificial neural network methods to forecast short-term solar power generation: a case study in Eastern Mediterranean Region

Solar power forecasting is substantial for the utilization, planning, and designing of solar power plants. Global solar irradiation (GSI) and meteorological variables have a crucial role in solar power generation. The ever-changing meteorological variables and imprecise measurement of GSI raise difficulties for forecasting photovoltaic (PV) output power. In this context, a major motivation appears for the accurate forecast of GSI to perform effective forecasting of the short-term output power of a PV plant. The presented study comprises of four artificial neural network (ANN) methods; recurrent neural network (RNN) method, feedforward backpropagation neural network (FFBPNN) method, support vector regression (SVR) method, and long short-term memory (LSTM) for daily total GSI prediction of Tarsus by using meteorological data. Moreover, this study proposes a model that utilizes the predicted daily GSI for output power forecasting of a grid-connected PV plant. The obtained results are compared with the output power generation data of a 350 kW solar power plant. The results are evaluated with the performance indices as mean absolute percentage error (MAPE), normalized root mean squared error (NRMSE), weighted mean absolute error (WMAE), and normalized mean absolute error (NMAE). FFBPNN method is chosen with the best results of MAPE 7.066%, NMAE 3.629%, NRMSE 4.673%, and WMAE 5.256%.

Evaluation of energy extraction of PV systems affected by environmental factors under real outdoor conditions

The global agenda to increase the renewable energy share has driven many countries and entities to harness solar energy from solar photovoltaic (PV) systems. However, the power generation of PV systems is strongly affected by climate conditions. Therefore, the main objective of this study is to analyze and predict the power generation of different PV technologies under arid desert climate conditions on an hourly basis. Two areas have been considered as case studies: Adrar in Algeria and Alice Springs in Australia. A total of nine physical models and input parameter combinations from six different power plants have been used and tested for the suitability of the proposed models for predicting the power yield of PV power plants depending on solar irradiance and other meteorological variables. Then, an ensemble learning technique is applied to improve the performance capabilities of the best-fit input combinations. The results reveal that the global irradiance, ambient air temperature, and relative humidity combination are the most related to the PV power output of all technologies under all-sky conditions and provide effective and efficient performance with the proposed ensemble learning, with an estimated accuracy of over 99%.

Country-wide solar power load profile for Germany 2015 to 2019: The impact of system curtailments on prediction models

A large (43824 h) country-wide solar power load profile (LP), solar irradiance and meteorological dataset (ten variables) for Germany covering years 2015 to 2019 is compiled and forecast with eight machine learning and deep learning (ML/DL) algorithms. Analysis reveals that system curtailments are likely responsible for some outlying predictions. The adaptive boosting (ADA) and random forest (RF) algorithms outperform convolutional neural networks and other algorithms in supervised prediction and forecasting tasks with the dataset. Once tuned with 2015 to 2018 data, ADA and RF forecast 2019 hourly data with root mean squared error (RMSE) of < 0.03. Similar forecasting accuracy is achieved using smaller datasets of three months of historical data from 2015 to 2018 to forecast hourly LP in each month of 2019. Forecasts for July 2019 are associated with the highest errors (RMSE = 0.41). The transparent open box (TOB) algorithm, because it can reveal details of data matching contributions to each of its forecasts, is used to data mine the July 2019 forecasts and conduct in-depth outlier analysis. It reveals nine extreme outliers that each overestimate LP based on irradiance and meteorological inputs. Analysis suggests that substantial overestimates leading to high RMSE for July 2019 are likely due to system curtailments. It is beneficial to combine less transparent (e.g., ADA and RF) with more transparent (e.g. TOB) ML algorithms to accurately forecast and data mine large solar power data sets. It also indicates that solar LP forecasts cannot always rely on irradiance and meteorological variables in isolation. There is a need to be mindful of system constraints and market conditions when predicting LP on a country-wide basis.

Solar Irradiance Nowcasting System Trial and Evaluation for Islanded Microgrid Control Purposes

The rapid increase in solar photovoltaic (PV) integration into electricity networks introduces technical challenges due to varying PV outputs. Rapid ramp events due to cloud movements are of particular concern for the operation of remote islanded microgrids (IMGs) with high solar PV penetration. PV systems and optionally controllable distributed energy resources (DERs) in IMGs can be operated in an optimised way based on nowcasting (forecasting up to 60 min ahead). This study aims to evaluate the performance under Perth, Western Australian conditions, of an all-sky imager (ASI)-based nowcasting system, installed at Murdoch University in Perth, Western Australia (WA). Nowcast direct normal irradiance (DNI) and global horizontal irradiance (GHI) are inputted into a 5 kWp solar PV system with a direct current (DC) power rating/alternating current (AC) power rating ratio of 1.0. A newly developed classification method provided a simplified irradiance variability classification. The obtained nowcasting system evaluation results show that the nowcasting system’s accuracy decreases with an increase in lead time (LT). Additionally, the nowcasting system’s accuracy is higher when the weather is either mostly clear (with a recorded LT15 mean absolute deviation (MAD) of 0.38 kW) or overcast (with a recorded LT15 MAD of 0.19 kW) than when the weather is intermittently cloudy with varying cloud conditions (with a recorded LT15 MAD of 0.44 kW). With lower errors observed in lower LTs, overall, it might be possible to integrate the nowcasting system into the design of IMG controllers. The overall performance of the nowcasting system at Murdoch University was as expected as it is comparable to the previous evaluations in five other different sites, namely, PSA, La Africana, Evora, Oldenburg, and Julich.

A machine learning methodology for the generation of a parameterization of the hydroxyl radical

Abstract. We present a methodology that uses gradient-boosted regression trees (a machine learning technique) and a full-chemistry simulation (i.e., training dataset) from a chemistry–climate model (CCM) to efficiently generate a parameterization of tropospheric hydroxyl radical (OH) that is a function of chemical, dynamical, and solar irradiance variables. This surrogate model of OH is designed to be integrated into a CCM and allow for computationally efficient simulation of nonlinear feedbacks between OH and tropospheric constituents that have loss by reaction with OH as their primary sinks (e.g., carbon monoxide (CO), methane (CH4), volatile organic compounds (VOCs)). Such a model framework is advantageous for studies that require multi-decadal simulations of CH4 or multi-year sensitivity simulations to understand the causes of trends and variations of CO and CH4. To allow the user to easily target the training dataset towards a desired application, we are outlining a methodology to generate a parameterization of OH and not presenting an “off-the-shelf” version of a parameterization to be incorporated into a CCM. This provides for the relatively easy creation of a new parameterization in response to, for example, changes in research goals or the underlying CCM chemistry and/or dynamics schemes. We show that a sample parameterization of OH generated from a CCM simulation is able to reproduce OH concentrations with a normalized root-mean-square error of approximately 5 % and capture the global mean methane lifetime within approximately 1 %. Our calculated accuracy of the parameterization assumes inputs being within the bounds of the training dataset. Large excursions from these bounds will likely decrease the overall accuracy. However, we show that the sample parameterization predicts large deviations in OH for an El Niño event that was not part of the training dataset and that the spatial distribution and strength of these deviations are consistent with the event. This result gives confidence in the fidelity of a parameterization developed with our methodology to simulate the spatial and temporal responses of OH to perturbations from large variations in the chemical, dynamical, and solar irradiance drivers of OH. In addition, we discuss how two machine learning metrics, Gain feature importance and Shapley additive explanations values, indicate that the behavior of a parameterization of OH generally accords with our understanding of OH chemistry, even though there are no physics- or chemistry-based constraints on the parameterization.

Mapping an Observation-Based Global Solar Irradiance Climatology across the Conterminous United States

Abstract The exponential growth in solar radiation measuring stations across the conterminous United States permits the generation of gridded solar irradiance data that capture the spatiotemporal variability of solar irradiance far more accurately than was previously possible from ground-based observations. Taking advantage of these observations, we generated a 30-yr climatology (1991–2020) of mean monthly global irradiance at a resolution of 30 arc s (∼800 m) on both a horizontal surface and a sloped ground surface. This paper describes the methods used to generate the gridded data, which include extensive quality control of station data, spatial interpolation of effective cloud transmittance using the “PRISM” method, and simulation of the effects of elevation, shading, and reflection from nearby terrain on solar irradiance. A comparison of the new dataset with several other solar radiation products reveals some spatial features in solar radiation that are either lacking or underresolved in some or all of the other datasets. Examples of these features include strong gradients near foggy coastlines and along mountain ranges where there is persistent orographically driven cloud formation. The workflow developed to create the long-term means will be used as a template for generating time series of monthly and daily solar radiation grids up to the present.

One-Day-Ahead Solar Irradiation and Windspeed Forecasting with Advanced Deep Learning Techniques

In recent years, demand for electric energy has steadily increased; therefore, the integration of renewable energy sources (RES) at a large scale into power systems is a major concern. Wind and solar energy are among the most widely used alternative sources of energy. However, there is intense variability both in solar irradiation and even more in windspeed, which causes solar and wind power generation to fluctuate highly. As a result, the penetration of RES technologies into electricity networks is a difficult task. Therefore, more accurate solar irradiation and windspeed one-day-ahead forecasting is crucial for safe and reliable operation of electrical systems, the management of RES power plants, and the supply of high-quality electric power at the lowest possible cost. Clouds’ influence on solar irradiation forecasting, data categorization per month for successive years due to the similarity of patterns of solar irradiation per month during the year, and relative seasonal similarity of windspeed patterns have not been taken into consideration in previous work. In this study, three deep learning techniques, i.e., multi-head CNN, multi-channel CNN, and encoder–decoder LSTM, were adopted for medium-term windspeed and solar irradiance forecasting based on a real-time measurement dataset and were compared with two well-known conventional methods, i.e., RegARMA and NARX. Utilization of a walk-forward validation forecast strategy was combined, firstly with a recursive multistep forecast strategy and secondly with a multiple-output forecast strategy, using a specific cloud index introduced for the first time. Moreover, the similarity of patterns of solar irradiation per month during the year and the relative seasonal similarity of windspeed patterns in a timeseries measurements dataset for several successive years demonstrates that they contribute to very high one-day-ahead windspeed and solar irradiation forecasting performance.

Influence of Aerosol Embedded in Shallow Cumulus Cloud Fields on the Surface Solar Irradiance

AbstractUbiquitous shallow cumulus clouds are associated with complex variability in surface solar irradiance (SSI). Aerosol embedded in the cloud field typically has a much smaller overall radiative effect, but can significantly perturb the shape of the SSI probability density function (PDF). These perturbations have important implications for several applications that utilize SSI, but are poorly quantified and are the subject of this study. Multiple cases of shallow cumulus cloud fields with embedded aerosol are simulated at the Southern Great Plains Atmospheric Observatory using large eddy simulation (LES). The LES‐derived cloud and aerosol fields are then ingested into Monte Carlo three‐dimensional (3D) radiative transfer to simulate SSI. We find a variety of perturbations to the SSI PDF that depend on aerosol presence and optical properties. The processes leading to these perturbations include extinction of the direct beam that often increases from the clear‐sky region toward cloud edge due to aerosol hygroscopic growth, and scattering of radiation by aerosol into cloud shadows. The ability to predict the SSI PDF in the presence of aerosol is assessed by adding three representative aerosol optical properties into an existing machine learning framework. We show that machine learning accurately predicts the SSI PDF across a wide range of conditions with negligible computational expense. Importance metrics reveal the relatively high influence of aerosol optical properties in making the predictions. These new findings highlight the important role that aerosol plays in SSI variability for highly 3D cloud‐aerosol environments and provides a computationally efficient route forward for its simulation.

Long-Term Trend Analysis in the Solar Radiation and Climate Experiment (SORCE)/Spectral Irradiance Monitor (SIM)

The Solar Radiation and Climate Experiment/Spectral Irradiance Monitor (SORCE/SIM) instrument was launched on 25 January 2003 with mission termination occurring on 25 February 2020. The SORCE/SIM provides a unique data set of the variability in solar spectral irradiance (SSI) during the descending phase of Solar Cycle 23 (SC23) from April 2003 to February 2009, the weaker solar-maximum conditions of SC24, and the quiescent SC24/SC25 minimum. The determination of the magnitude and phase of SSI variations rely on the unambiguous determination of the effects of the space environment and solar-exposure-related degradation mechanisms. The instrument-only corrections for SIM are based on a comparison of two functionally identical (mirror image) prism spectrometers with four independent detectors in each spectrometer channel. The degradation correction is strictly instrumental in its methodology and makes no assumptions about the magnitude, slope, or wavelength dependence of the SSI variability.

Climate Change Impact Assessment for Future Wind and Solar Energy Installations in India

Solar and wind assets are climate-dependent and changes in climate will result in variations in their generation and intermittency. Developers of solar and wind parks in India have observed changes in climate conditions and variability in solar irradiation and wind profiles at the seasonal and year-to-year timescales. Future climate change perturbations, including monsoon shifts, could lead to lower-than-predicted wind and solar energy production and affect the economics of solar and wind assets. Regional climate models (RCMs) are the basis of climate impact assessments and the most trusted source of information to extract knowledge about future trends in climate variables. However, RCM projections are tainted with variability and uncertainty about the future trends. For India as a case study, we use the RCMs generated by the Coordinated Regional Climate Downscaling Experiment West Asia project (CORDEX WAS) to calculate individual wind, radiation, and temperature trends at selected sites; estimate wind and solar PV energy time series; and embed them in portfolio methods to test the impact of combining wind and solar assets on the variability of the total production and the uncertainty about the predicted production. We include a comparison of CORDEX RCMs with the ERA5 reanalysis dataset and conclude that all available RCMs reasonably simulate the main annual and seasonality features of wind speed, surface solar radiation, and temperature in India. The analysis demonstrates that the uncertainty about the portfolio return can be reduced by optimizing the combination of wind and solar assets in a producer portfolio, thus mitigating the economic impact of climate change. We find that the reduction obtained with a mixed portfolio ranges from 33 to 50% compared to a wind only portfolio, and from 30 to 96% compared to a solar only portfolio.

Grey wolf optimization-recurrent neural network based maximum power point tracking for photovoltaic application

To increase the photovoltaic (PV) power-generation conversion, MPPT is the primary concern. This works explains about the grey wolf optimization (GWO - RNN)-based hybrid maximum power point tracking (MPPT) method to get quick and maximum photovoltaic (PV) power with zero oscillation tracking. The GWO – RNN based MPPT method doesn’t need additional sensor for measuring irradiance and temperature variables. The NLT is used for the multi-level inverter (MLI) control strategy to achieve less harmonics distraction and less switching losses with better voltage and current profile. This employed methodology brings remarkable aspects in the PV boosting potential extraction. A GWO – RNN controlled LUO converter is a zero output harmonic agreement impedance matching interface that is MPPT is performed by placing the PV modules between the load regulator power circuit and the load regulator power circuit. To actualize the proposed hybrid GWO – RNN model for the PV system, perturb and observe, RNN, ant colony optimization, and artificial bee colony MPPT techniques are employed. The MATLAB interfaced dSPACE interface is used to finish the hands-on validation of the intended grid-integrated PV system. The obtained results eloquently support the appropriate design of higher-performance control algorithms.

Temperature and irradiance variability on the Egyptian coast of the Red Sea during the 2020 bleaching event

Temperature and irradiance variability on the Egyptian coast of the Red Sea during the 2020 bleaching event

High temporal variability of surface solar irradiance due to cloud enhancement effect over the Western Ghat mountains in peninsular India

Western Ghat (WG) mountains along the west coast of peninsular India play an important role in the distribution of monsoon rainfall on its windward and leeward sides. The study unravels the increase in cloudiness and also increase in the Cloud Enhancement of Solar irradiance (CES) during the dry and wet seasons in the windward (high altitude) and the leeward (low altitude) sides of WG. From May 2018 to May 2021, the global horizontal irradiance (GHI) reveals wide temporal and spatial variability with a high frequency of CES events, even exceeding the extraterrestrial value around the noontime during the wet season (May to October). The observed excess irradiance due to partial/transparent clouds compared to clear sky maxima is more than 400 W/m2 on the windward and 350 W/m2 on the leeward. With the increase in cloudy days, diffused irradiance also increased with greater values during CES events. CES induced positive shortwave cloud radiative forcing (CRF) varied up to +400 W/m2 (warming effect), while negative CRF varied up to −1000 W/m2 (cooling effect). During monsoon, the warming due to frequent CES events could locally offset a large cooling to some extent. Clouds resulting in huge irradiance variability in a day could also induce large fluctuations in renewable solar energy. Such short-term sporadic local events are averaged out in satellite data.

Short-Term Solar Power Predicting Model Based on Multi-Step CNN Stacked LSTM Technique

Variability in solar irradiance has an impact on the stability of solar systems and the grid’s safety. With the decreasing cost of solar panels and recent advancements in energy conversion technology, precise solar energy forecasting is critical for energy system integration. Despite extensive research, there is still potential for advancement of solar irradiance prediction accuracy, especially global horizontal irradiance. Global Horizontal Irradiance (GHI) (unit: KWh/m2) and the Plane Of Array (POA) irradiance (unit: W/m2) were used as the forecasting objectives in this research, and a hybrid short-term solar irradiance prediction model called modified multi-step Convolutional Neural Network (CNN)-stacked Long-Short-Term-Memory network (LSTM) with drop-out was proposed. The real solar data from Sweihan Photovoltaic Independent Power Project in Abu Dhabi, UAE is preprocessed, and features were extracted using modified CNN layers. The output result from CNN is used to predict the targets using a stacked LSTM network and the efficiency is proved by comparing statistical performance measures in terms of Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Squared Error (MAE), and R2 scores, with other contemporary machine learning and deep-learning-based models. The proposed model offered the best RMSE and R2 values of 0.36 and 0.98 for solar irradiance prediction and 61.24 with R2 0.96 for POA prediction, which also showed better performance as compared to the published works in the literature.

A multi-instrument investigation of the frequency stability of oscillations above the acoustic cut-off frequency with solar activity

ABSTRACT Below the acoustic cut-off frequency, oscillations are trapped within the solar interior and become resonant. However, signatures of oscillations persist above the acoustic cut-off frequency, and these travelling waves are known as pseudo-modes. Acoustic oscillation frequencies are known to be correlated with the solar cycle, but the pseudo-mode frequencies are predicted to vary in antiphase. We have studied the variation in pseudo-mode frequencies with time systematically through the solar cycle. We analysed Sun-as-a-star data from Variability of Solar Irradiance and Gravity Oscillations (VIRGO), and Global Oscillations at Low Frequencies (GOLF), as well as the decomposed data from Global Oscillation Network (GONG) for harmonic degrees 0 ≤ l ≤ 200. The data cover over two solar cycles (1996–2021, depending on instrument). We split them into overlapping 100-d long segments and focused on two frequency ranges, namely 5600–$6800\, \rm \mu Hz$ and 5600–$7800\, \rm \mu Hz$. The frequency shifts between segments were then obtained by fitting the cross-correlation function between the segments’ periodograms. For VIRGO and GOLF, we found no significant variation of pseudo-mode frequencies with solar activity. However, in agreement with previous studies, we found that the pseudo-mode frequency variations are in antiphase with the solar cycle for GONG data. Furthermore, the pseudo-mode frequency shifts showed a double-peak feature at their maximum, which corresponds to solar activity minimum, and is not seen in solar activity proxies. An, as yet unexplained, pseudo-periodicity in the amplitude of the variation with harmonic degree l is also observed in the GONG data.

Spatio-temporal characterization of long-term solar resource using spatial functional data analysis: Understanding the variability and complementarity of global horizontal irradiance in Ecuador

Understanding the spatio-temporal variability of the solar resource is crucial to effectively support solar power utilization. Unfortunately, long-term and high-resolved measurements of solar irradiance are generally scarce, challenging the characterization for larger areas. In this paper, we propose a methodology to characterize the spatio-temporal variability of global horizontal irradiance (GHI) at a regional scale using long-term satellite-derived data. Spatial functional data analysis (sFDA) is used to identify areas with similar intra-annual variability patterns. The methodology is applied to a 21-year period data on Ecuador retrieved from the National Solar Radiation Database. Being the first time that sFDA is used for this purpose, the results indicate that it provides an appropriate basis for the interannual variability and complementarity analyses. In Ecuador's mainland, twenty-two subregions with four seasonal patterns are identified. The highest GHI potential (5.4 kWhm−2d−1) with the lowest variability (3.4%) is found in the Inter-Andean valleys. Further, seasonal complementarities between the coast and western Andes are identified. In Galapagos, high values are found over all islands (≥4.8 kWhm−2d−1), characterized by three subregions with one seasonal pattern. Our findings provide the first comprehensive spatio-temporal characterization of GHI in Ecuador, which aims at supporting a sustainable energy transition in the country.

Irradiance Nowcasting by Means of Deep-Learning Analysis of Infrared Images

This work proposes and evaluates a method for the nowcasting of solar irradiance variability in multiple time horizons, namely 5, 10, and 15 min ahead. The method is based on a Convolutional Neural Network structure that exploits infrared sky images acquired through an All-Sky Imager to estimate the range of possible values that the Clear-Sky Index will possibly assume over a selected forecast horizon. All data available, from the infrared images to the measurements of Global Horizontal Irradiance (necessary in order to compute Clear-Sky Index), are acquired at SolarTechLAB in Politecnico di Milano. The proposed method demonstrated a discrete performance level, with an accuracy peak for the 5 min time horizon, where about 65% of the available samples are attributed to the correct range of Clear-Sky Index values.