Solar Radiation Time Series Research Articles

Stochastic weather generator for the design and reliability evaluation of desalination systems with Renewable Energy Sources

Abstract The operation of Renewable Energy Sources (RES) systems is highly affected by the continuously changing meteorological conditions and the design of a RES system has to be robust to the unknown weather conditions that it will encounter during its lifetime. In this paper, the use of Stochastic Weather Generators (SWGENs) is introduced for the optimal design and reliability evaluation of hybrid Photovoltaic/Wind-Generator systems providing energy to desalination plants. A SWGEN is proposed, which is based on parametric Markov-Switching Auto-Regressive (MSAR) models and is capable to simulate realistic hourly multivariate time series of solar irradiance, temperature and wind speed of the target installation site. Numerical results are presented, demonstrating that: (i) SWGENs enable to evaluate the reliability of RES-based desalination plants during their operation over a 20 years lifetime period and (ii) using an appropriate time series simulated with a SWGEN as input to the design optimization process results in a RES-based desalination plant configuration with higher reliability compared to the configurations derived when the other types of meteorological datasets are used as input to the design optimization process.

Estimation of solar radiation using stepwise multiple linear regression with principal component analysis in Algeria

For the modelling of solar systems, reliable and complete time series of solar radiation are required. However, solar radiation data from ground measurements or satellite images may be available only for limited time periods and often have data quality issues or/and data gaps. It is possible to get reasonably accurate radiation estimates of solar radiation from meteorological parameters, which can complement or extend existing data. In this paper, a model based on eight meteorological parameters (evaporation, temperature, wind speed, visibility, cloud cover, sunshine duration, sunshine ratio and global solar radiation) is developed to predict the global solar radiation using 29-year (1986–2014) data from Oran radiometric station in Algeria. Two versions of the multiple regression analysis are used: the first with original observation variables (called as manifest variables) and the second by PCs called as latent variables or (common) factors. Separate analyses have been carried out for two scenarios: overall and partial study. The first scenario uses all climatic dataset, but the last scenario uses three subsets which are sunny, partly cloudy and cloudy days. This classification has been obtained using certain statistical properties that we have considered as thresholds. Stepwise multiple linear regression analysis is used to fit global solar radiation data using meteorological variables as predictors. A variable selection method based on PCA technique are used to obtain the subsets of predictors to be included in the regression model of global solar radiation data. Hence, the proposed relationships may be considered useful for predicting global solar radiation rate in other sites of Algeria that have climatic conditions similar to the study area.

Clustering methods assessment for investment in zero emission neighborhoods’ energy system

This paper investigates the use of clustering in the context of designing the energy system of Zero Emission Neighborhoods (ZEN). ZENs are neighborhoods who aim to have net zero emissions during their lifetime. While previous work has used and studied clustering for designing the energy system of neighborhoods, no article dealt with neighborhoods such as ZEN, which have high requirements for the solar irradiance time series, include a CO2 factor time series and have a zero emission balance limiting the possibilities. To this end several methods are used and their results compared. The results are on the one hand the performances of the clustering itself and on the other hand, the performances of each method in the optimization model where the data is used. Various aspects related to the clustering methods are tested. The different aspects studied are: the goal (clustering to obtain days or hours), the algorithm (k-means or k-medoids), the normalization method (based on the standard deviation or range of values) and the use of heuristic. The results highlight that k-means offers better results than k-medoids and that k-means was systematically underestimating the objective value while k-medoids was constantly overestimating it. When the choice between clustering days and hours is possible, it appears that clustering days offers the best precision and solving time. The choice depends on the formulation used for the optimization model and the need to model seasonal storage. The choice of the normalization method has the least impact, but the range of values method show some advantages in terms of solving time. When a good representation of the solar irradiance time series is needed, a higher number of days or using hours is necessary. The choice depends on what solving time is acceptable.

Spatiotemporal variability and trends in global solar radiation over Poland based on satellite‐derived data (1986–2015)

AbstractIncoming solar radiation is the most important factor shaping climate system on Earth and the main element of the surface heat balance. The main aim of this study was to investigate the changes in the amount of global solar radiation reaching the Earth's surface in Poland during the 30‐year period 1986–2015. Trends in changes and fluctuations in the size of global solar radiation over Poland were determined. The solar radiation was described based on satellite products originating from the Surface Incoming Shortwave Radiation product from the Surface Solar Radiation Data Set – Heliosat, Edition 2 (SARAH‐2). The average annual sum of global solar radiation over Poland amounted to 3,902 MJ·m−2. The average annual radiation sums were the smallest in northern Poland and mountain basins, while they were the largest in southern Poland. The average annual radiation sum over Poland increased by 7.16 MJ·m−2·year−1 on average. The areas with the largest increase in the amount of solar radiation had the smallest average radiation sums during the multi‐year period (Pomerania, Northern Poland), and those where the increase in radiation was moderate had the highest average radiation sums (Central and Southern Poland). This shows that the spatial differentiation of the amount of solar radiation over Poland was gradually decreasing during this period. A several‐year cycle (of 12–13 years) of annual fluctuations in global solar radiation sums was observed using wavelet analysis. The cycle was visible between the early 1990s and 2005. It resulted from the medium‐term cyclical component (an 11.3‐year cycle which was the strongest until 2010) that occurred in summer. In the long term, the occurrence of cycles in the time series of solar radiation may result from cyclical or quasi‐cyclical changes in aerosol concentration, but this requires a separate study and further in‐depth research based on much longer data series.

Inflection point in the power spectrum of stellar brightness variations

Context. Young and active stars generally have regular, almost sinusoidal, patterns of variability attributed to their rotation, while the majority of older and less active stars, including the Sun, have more complex and non-regular light curves, which do not have clear rotational-modulation signals. Consequently, the rotation periods have been successfully determined only for a small fraction of the Sun-like stars (mainly the active ones) observed by transit-based planet-hunting missions, such as CoRoT, Kepler, and TESS. This suggests that only a small fraction of such systems have been properly identified as solar-like analogues. Aims. We aim to apply a new method of determining rotation periods of low-activity stars, such as the Sun. The method is based on calculating the gradient of the power spectrum (GPS) of stellar brightness variations and identifying a tell-tale inflection point in the spectrum. The rotation frequency is then proportional to the frequency of that inflection point. In this paper, we compare this GPS method to already-available photometric records of the Sun. Methods. We applied GPS, auto-correlation functions, Lomb-Scargle periodograms, and wavelet analyses to the total solar irradiance (TSI) time series obtained from the Total Irradiance Monitor on the Solar Radiation and Climate Experiment and the Variability of solar IRradiance and Gravity Oscillations experiment on the SOlar and Heliospheric Observatory missions. We analysed the performance of all methods at various levels of solar activity. Results. We show that the GPS method returns accurate values of solar rotation independently of the level of solar activity. In particular, it performs well during periods of high solar activity, when TSI variability displays an irregular pattern, and other methods fail. Furthermore, we show that the GPS and light curve skewness can give constraints on facular and spot contributions to brightness variability. Conclusions. Our results suggest that the GPS method can successfully determine the rotational periods of stars with both regular and non-regular light curves.

Analysis of Artificial Neural Network Point Forecasting Models and Prediction Intervals for Solar Irradiance Estimation

An accurate knowledge on solar irradiance prediction is particularly required for proper development and planning of Photovoltaic (PV) energy systems. The main purpose of the present research is to assess the accuracy of Artificial Neural Networks (ANN) short-term forecast of univariate solar irradiance time series, with conventional point prediction and Prediction Intervals (PIs), comparing models. The Lower Upper Bound Estimation trained with Particle Swarm Optimization (PSO-LUBE) was used for PIs estimation. Solar irradiance data collected from a station in Amazon region in Brazil was used to train and test the models. Results demonstrate that all ANN models yield good accuracy in terms of prediction error: 8.1-8.5% for normalized root Mean Square Error (nRMSE), 5.8-6.0% for normalized Mean Absolute Error (nMAE) and 94-95% for determination coefficient (R2). However, due to the accuracy of PI information (Coverage Probability = 94.94% and PI Normalized Average Width = 32.50%), PSO-LUBE was the best method tested for decision-making.

A new artificial multi-neural approach to estimate the hourly global solar radiation in a semi-arid climate site

In this study, we propose a new hybrid machine-learning algorithm called artificial multi-neural approach, for developing a multi-input single-output (MISO) model to estimate the hourly global solar radiation (HGSR) time series in Agdal site (latitude 31° 37′ N, longitude 08° 01′ W, elevation 466 m), Marrakesh, Morocco. To achieve this goal, three training algorithms (scaled conjugate gradient, Levenberg-Marquardt, and resilient backpropagation) are selected to train the developed model, using some accessible hourly meteorological variables recorded during 7 years (from 2008 to 2014) as exogenous inputs. Furthermore, a pertinence determination test (PDT) is used to point out the most pertinent inputs for the accurate HGSR estimation. The best configuration of the multi-neural model includes five known input parameters as the best scenario, i.e., acquisition hour, air temperature, relative humidity, wind speed, and precipitation. Some statistical indicators are used to evaluate the performance of the developed model. The obtained results demonstrate the reliability and the precision of our proposed machine-learning algorithm, representing a good solution for estimating solar radiation time series needed to design and manage solar energy systems compared to the present standards.

Data-Driven Short-Term Solar Irradiance Forecasting Based on Information of Neighboring Sites

Short-term forecasting of solar irradiance at a targeted site with consideration of its time series and measurements at neighboring sites is studied in this paper. A data-driven framework for forecasting solar irradiance based on fusing spatial and temporal information is proposed. In the framework, data-driven approaches including boosted regression trees (BRT), artificial neural network, support vector machine, and least absolute shrinkage and selection operator are applied to model spatial dependence among solar irradiance time series thereby to generate forecasts of solar irradiance at the targeted site. A comprehensive comparison among data-driven forecasting models is performed using 30-min averaged data of recent two years. Moreover, benchmarking models including scaled persistence model, autoregressive model, and autoregressive exogenous model are employed to further validate the effectiveness of data-driven forecasting models. Computational results of multiple-steps ahead forecasting demonstrate that the BRT model gives the best performance with the lowest normalized root mean squared error of 18.4%, 24.3%, 27.9%, and 30.6% for forecasting horizons of 30, 60, 90, and 120-min, respectively.

SORCE‐Based Solar Spectral Irradiance (SSI) Record for Input Into Chemistry‐Climate Studies

AbstractWhole atmosphere chemistry‐climate model (CCM) transient studies require a solar spectral irradiance (SSI) time series with very broad wavelength coverage with an uninterrupted daily cadence over a full solar cycle time period. Herein is described the construction of a SSI record valid over Solar Cycles 23 and 24 (SC23 and SC24) specifically designed to provide: (1) a daily solar spectrum from SORCE (Solar Radiation and Climate Experiment) observations from 208–2400 nm over the 2003 to 2015 time frame, (2) a wavelength extension of this spectrum out to 10000 nm from daily spectral synthesis from the Solar Radiation Physical Model (SRPM), (3) SRPM gap‐filled spectra during the extended SORCE safe‐hold time period (Aug 2013–March 2014) and back to Solar Cycle 23 maximum conditions in 2001 and 2002, (4) analysis of solar images generated by the Precision Solar Photometric Telescope (PSPT) network from the Mauna Loa Solar Observatory (MLSO) and Rome Observatory (OAR) (These images act both as input to the SRPM model and as a stand‐alone proxy of solar variability.), (5) documentation of uncertainties and biases in the time series as a function of wavelength, and (6) comparative analysis of the SORCE time series with SRPM over the full wavelength range of the data set.

A new perspective on the sunshine duration variability

The sunshine duration data is analysed here from a new perspective, different from the traditional ones. The study is focused on pairs of periods of shining sun (called clear periods) followed by periods with the sun covered by clouds (called dark periods). Their statistical and sequential properties are illustrated by using results obtained from satellite observations and ground measurements at Timisoara (Romania). Solar irradiance time series measured at ground are converted into a database consisting of pairs of clear/dark periods. These pairs are associated with three cloud classes derived from satellite observations: C1 (the sky is cloud-free), C2 (cloudy sky) and C3 (sky with semi-transparent or fractional clouds). The study was conducted by using tools from survival analysis. This is a novelty in sunshine duration analysis. Since information is missing during the night, the clear/dark period correlation was studied considering a censoring procedure (for all nine combinations of cloud types at the beginning of dark and clear intervals). Cox regression has been used to study the influence of covariates (such as extraterrestrial solar irradiance and sun elevation angle) on sunshine duration distribution. Results show that increase of extraterrestrial solar radiation tends to increase the risk of stopping any clear or dark interval, i.e. it tends to increase the variability in solar irradiance. Regardless of cloud class, stochastically shorter intervals are found for increasing sun elevation angle.

Analyzing territory for the sustainable development of solar photovoltaic power using GIS databases

Solar energy generated by grid-connected photovoltaic (GCPV) systems is considered an important alternative electric energy source because of its clean energy production system, easy installation, and low operating and maintenance costs. This has led to it becoming more popular compared with other resources. However, finding optimal sites for the construction of solar farms is a complex task with many factors to be taken into account (environmental, social, legal and political, technical-economic, etc.), which classic site selection models do not address efficiently. There are few studies on the criteria that should be used when identifying sites for solar energy installations (large grid-connected photovoltaic systems which have more than 100 kWp of installed capacity). It is therefore essential to change the way site selection processes are approached and to seek new methodologies for location analysis. A geographic information system (GIS) is a tool which can provide an effective solution to this problem. Here, we combine legal, political, and environmental criteria, which include solar radiation intensity, local physical terrain, environment, and climate, as well as location criteria such as the distance from roads and the nearest power substations. Additionally, we use GIS data (time series of solar radiation, digital elevation models (DEM), land cover, and temperature) as further input parameters. Each individual site is assessed using a unique and cohesive approach to select the most appropriate locations for solar farm development in the Valencian Community, a Spanish region in the east of Spain.

Ensemble streamflow projections for a small watershed with HSPF model.

A watershed modeling tool, Hydrological Simulation Program-FORTRAN (HSPF), was utilized to model the hydrological processes in the agricultural Sarısu watershed in western Turkey. The meteorological input data were statistically downscaled time series from General Circulation Model simulations. The input data were constructed as an ensemble of 400 individual time series of temperature, precipitation, dewpoint temperature, solar radiation, potential evapotranspiration, cloudiness, and wind velocity, as required by HSPF. The ensemble was divided into four subsets, each comprising of 100 time series, of different Special Report on Emissions Scenarios. Yearly and monthly total streamflow time series were obtained from the calibrated and validated HSPF model spanning a period of 116 years between the water years of 1984 and 2099. The projections in the watershed showed a median increase of 3 °C in yearly average temperatures between the beginning and end 30-year periods of the 116-year simulation periods based on 400 ensemble members while the corresponding change in total yearly precipitation was - 71 mm. These changes led to a decrease in yearly streamflows by 40% which reflected itself to varying degrees in monthly flows. Correlations were established between the principal drivers of the watershed hydrological cycle, namely temperature and precipitation, and streamflow. The results showed that the changes in the climatic conditions will greatly affect water-related issues in the watershed and emphasize the necessity of preparing carefully to adapt to a warmer and drier climate.

Solar irradiance time series derived from high-quality measurements, satellite-based models, and reanalyses at a near-equatorial site in Brazil

This study analyzes five years of 1-min solar global horizontal irradiance (GHI) and direct normal irradiance (DNI) observations obtained at Petrolina (northeast Brazil). Quality-assured hourly and daily averages are obtained after applying filters and methodologies based on a Baseline Solar Radiation Network (BSRN) quality-control procedure. To calculate correct hourly averages, a minimum fraction of 20% of valid GHI or DNI minutely data is needed, as well as at least 60% of valid days to calculate correct daily-mean monthly values. An asymmetric diurnal pattern is found in GHI and DNI during all months, attributed to consistently higher cloudiness in the morning. The quality-assured hourly and monthly-mean GHI and DNI time series are compared to estimates from 11 solar databases regularly used in solar resource assessment studies: CAMS, CERES, ERA5, INPE, MERRA-2, Meteonorm, NASA-POWER, NSRDB, SARAH, SWERA-BR, and SWERA-US. For hourly GHI values, a range of RMS differences is found between the best (CAMS, 17.3%) and the worst (MERRA-2, 38.9%) results. The latter database is also affected by a larger bias (18.7%) than CAMS (4%). Larger RMS differences are found with hourly DNI, in a range extending from 37% (CAMS) to 63.4% (ERA5). Biases are all above 12%, except for CERES (−1%). For long-term mean-monthly GHI results, low biases of less than 1% are obtained with CAMS, CERES and NASA-POWER, whereas MERRA-2 overestimates (13%). Larger biases are found for mean-monthly DNI, spanning between CAMS (3%) and Meteonorm (−18.4%). Overall, CAMS appears the most consistent solar database for long-term irradiance time series at Petrolina. The significant differences found here between modeled databases are larger than expected, and underline the importance of regional validation studies like this one to decrease the incidence of uncertainties in solar resource assessments on the design and performance of solar energy projects.

Algorithmic Probability Method Versus Kolmogorov Complexity with No-Threshold Encoding Scheme for Short Time Series: An Analysis of Day-To-Day Hourly Solar Radiation Time Series over Tropical Western Indian Ocean.

The complexity of solar radiation fluctuations received on the ground is nowadays of great interest for solar resource in the context of climate change and sustainable development. Over tropical maritime area, there are small inhabited islands for which the prediction of the solar resource at the daily and infra-daily time scales are important to optimize their solar energy systems. Recently, studies show that the theory of the information is a promising way to measure the solar radiation intermittency. Kolmogorov complexity (KC) is a useful tool to address the question of predictability. Nevertheless, this method is inaccurate for small time series size. To overcome this drawback, a new encoding scheme is suggested for converting hourly solar radiation time series values into a binary string for calculation of Kolmogorov complexity (KC-ES). To assess this new approach, we tested this method using the 2004–2006 satellite hourly solar data for the western part of the Indian Ocean. The results were compared with the algorithmic probability (AP) method which is used as the benchmark method to compute the complexity for short string. These two methods are a new approach to compute the complexity of short solar radiation time series. We show that KC-ES and AP methods give comparable results which are in agreement with the physical variability of solar radiation. During the 2004–2006 period, an important interannual SST (sea surface temperature) anomaly over the south of Mozambique Channel encounters in 2005, a strong MJO (Madden–Julian oscillation) took place in May 2005 over the equatorial Indian Ocean, and nine tropical cyclones crossed the western part of the Indian Ocean in 2004–2005 and 2005–2006 austral summer. We have computed KC-ES of the solar radiation time series for these three events. The results show that the Kolmogorov complexity with suggested encoding scheme (KC-ES) gives competitive measure of complexity in regard to the AP method also known as Solomonoff probability.

Prediction of time series for several hours of surface solar irradiance using one-granule cloud property data from satellite observations

Satellite observation is one of the most useful methods for monitoring the power generated by photovoltaic systems. We propose a method for predicting time series of the surface solar irradiance (SSI) from one-granule cloud properties data obtained by satellite observation. This method has two parts. One is prediction of time series features from cloud properties, which is based on a previous study. The other is prediction of a time series from time series features, which is developed in this work. Seven time series features are used: mean, standard deviation, skewness, kurtosis, linear regression coefficient, autocorrelation coefficient with lag-1, and sample entropy. The prediction skill is characterized in two ways: by the distance between the predicted and observed time series (metric D), and by the similarity of the spectrum between the two time series (metric S). Verifying the time series prediction method from time series features shows that the mean and linear regression coefficient are the first and second most influential factors for metric D, and standard deviation and autocorrelation coefficient with lag-1 are those for metric S.The prediction skill of the prediction method for the time series features from cloud properties is also verified, indicating that predictors for the mean, standard deviation lag-1 autocorrelation coefficient, and sample entropy have sufficient prediction skill, measured by the standardized root mean square error. Finally, the two methods are integrated to construct the prediction system using four time series features. The overall skill of the prediction system for the time series is evaluated. The proposed system provides information about the strength of the SSI time series and its variation for several hours over the satellite observation area.

Day-ahead spatiotemporal solar irradiation forecasting using frequency-based hybrid principal component analysis and neural network

Owing to a shortage of fossil fuels, environmental pollution and the greenhouse effect, renewable energy generation has become important in a modern smart grid. However, the characteristics of renewable power generation are volatile and uncertain. This work proposes a new day-ahead spatiotemporal forecasting method for solar irradiation to ensure the efficient operation of power systems. First, the frequency features of a solar irradiation time-series are extracted by discrete Fourier transform (DFT). Principal component analysis (PCA) is then used to identify the crucial frequency features, which are input to an Elman-based neural network to carry out subsequent 24-hour (day ahead) solar irradiation forecasting. Historical weather data from five weather stations that are near the target location are used. Comparative studies of traditional Autoregressive Integrated Moving Average Model (ARIMA), PCA-Back-Propagation (BP)-based neural network, the persistence method, DFT-PCA-BP and the proposed DFT-PCA-Elman method reveal that the proposed method is the most accurate in day-ahead forecasting using spatiotemporal data.

Nonparametric Bayesian-based recognition of solar irradiance conditions: Application to the generation of high temporal resolution synthetic solar irradiance data

High resolution synthetic irradiance is of interest for theoretical studies such as grid integration of solar PV and battery storage analysis. Access to site-specific data is often limited to inadequate temporal resolutions for such application. A new model for producing synthetic solar global horizontal irradiance (GHI) time-series at up to 1-min resolution is presented as derived from >10-min input data. Briefly, it is a clustered-based method for daily clearness index distributions using Dirichlet process Gaussian mixture model (DPGMM). DPGMM is a nonparametric Bayesian (NPB) model indexed with an infinite-dimensional space of parameters. The key benefit of the NPB paradigm is the automatic adaptation to the correct complexity level and model size, suggesting a local adaptation of the model to all climatic conditions. A posterior inference using Markov chain Monte Carlo algorithm (namely Gibbs sampling) is applied. The model only requires a valid number of intraday data to construct daily distributions, then it can be applied worldwide. The synthetic GHI time series are validated against observed 1-min GHI data for four locations distributed throughout the world with different climatic conditions and significant geographic separation. Moreover, the presented method can generate data based on similar climatic conditions. A good fit between real and generated data is observed. We present an nRMSE ≤4% and nMBE <±4% between generated and measured means at both daily and monthly scales for all sites. The agreement between the real and generated cumulative density distributions of six comparative variability metrics (defined in text) at four different sites is measured using the overlapping and the Kullback–Leibler coefficients, which are ≥75% and ≤10% respectively, in all cases. To ensure the reproducibility of the research presented in this paper, the methodology is freely available as an R-package downloadable from SolarClusGnr.

Assessing solar resource and photovoltaic production in Tahiti from ground-based measurements

This study focuses on the solar resource available at Faaa, Tahiti (17.5°S, 149.5°W) thanks to 10 year-long solar irradiance time series. Faaa’s global horizontal irradiance ranges from 14 MJ.m-2.day-1 (June) to 21 MJ.m-2.day-1 (November) in agreement with the sun’s annual path, while clearness index ranges from 0.5 (January) to 0.67 (July), in agreement with the wet and dry seasons. The Global Solar Atlas satellite-derived dataset shows acceptable relative error when compared to Faaa in situ measurements. This product could then be used for other coastal areas of Tahiti. The annual energy output of a single PV module is 256.7 kWh, which corresponds to 7 % of the annual consumption of a typical household in Tahiti. The capacity factor reaches 22.5 %, which makes Faaa a good site for harnessing solar resource.

Spatial and Temporal Non-Linear Dynamics Analysis and Predictability of Solar Radiation Time Series for La Reunion Island (France).

Analysis of daily solar irradiation variability and predictability in space and time is important for energy resources planning, development, and management. The natural intermittency of solar irradiation is mainly triggered by atmospheric turbulent conditions, radiative transfer, optical properties of cloud and aerosol, moisture and atmospheric stability, orographic and thermal forcing, which introduce additional complexity into the phenomenological records. To address this question for daily solar irradiation data recorded during the period 2011–2015, at 32 stations measuring solar irradiance on La Reunion French tropical Indian Ocean Island, we use the tools of non-linear dynamics: the intermittency and chaos analysis, the largest Lyapunov exponent, Sample entropy, the Kolmogorov complexity and its derivatives (Kolmogorov complexity spectrum and its highest value), and spatial weighted Kolmogorov complexity combined with Hamming distance to assess complexity and corresponding predictability. Finally, we have clustered the Kolmogorov time (that quantifies the time span beyond which randomness significantly influences predictability) for daily cumulative solar irradiation for all stations. We show that under the record-breaking 2011–2012 La Nina event and preceding a very strong El-Nino 2015–2016 event, the predictability of daily incident solar energy over La Réunion is affected.

Predictability of Characteristics of Temporal Variation in Surface Solar Irradiance Using Cloud Properties Derived from Satellite Observations

AbstractUnderstanding of the characteristics of variation in surface solar irradiance on time scales shorter than several hours has been limited because ground-based observation stations are located coarsely. However, satellite observation data can be used to bridge this gap. We propose an approach for predicting characteristics of a time series of surface solar irradiance in a 121-min time window for areas without ground-based measurement systems. Time series features—mean, standard deviation, and sample entropy—are used to represent the characteristics of variation in surface solar irradiance quantitatively. We examine cloud properties over the area to design prediction models of these time series features. Cloud properties averaged over the defined domain and texture features that represent characteristics of the spatial distribution of clouds are used as measures of cloud features. Predictors for time series features, where explanatory variables are cloud features, are constructed employing the random-forest regression method. The performance test for predictions indicates that the mean and standard deviation can be predicted with higher prediction skill, whereas the predictor for sample entropy has lower prediction skill. The importance of cloud features for predictors and partial dependence of the predictors on explanatory variables are also analyzed. Cloud optical thickness (COT) and cloud fraction (CFR) were important for predicting the mean. Two texture features—contrast and local homogeneity (LHM)—and COT were important for predicting the standard deviation, and COT, LHM, and CFR were important for predicting the sample entropy. These results indicate which satellite-derived cloud field properties are useful for predicting time series features of surface solar irradiance.