Generation of synthetic solar datasets for risk analysis

Larrañeta, M. G.; Fernández-Peruchena, Carlos M.; Silva-Pérez, M.A.; Lillo-Bravo, I.; Grantham, Adrian; Boland, John

doi:10.1016/j.solener.2019.05.042

Cited by 16 publications

(7 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This methodology allows the solar resource assessment-and thus the energy output calculation-to be performed in a way that is similar to that currently used for estimating other essential variables in the economic assessment of solar power plants. The generation of hundreds of such plausible years has been demonstrated by Larrañeta et al (2019), Fernández-Peruchena et al (2015, and Meybodi et al (2017). Other authors-e.g., Ho, Khalsa, Kolb (2011) and Ho and Kolb (2010)-have found issues with the Monte Carlo approach and suggested the Latin hypercube sampling method instead.…”

Section: Chapter 9-22mentioning

confidence: 97%

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications: Third Edition

Sengupta

Habte

Wilbert

et al. 2021

View full text Add to dashboard Cite

Section: Chapter 9-22mentioning

confidence: 97%

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications: Third Edition

Sengupta

Habte

Wilbert

et al. 2021

View full text Add to dashboard Cite

“…In the direction of the third strategy (i.e., using grey-or black-box modeling strategy), the most recent models are copula [26], neural-network based [27], ensemble methods [28], deep learning [29], and MC [30,31]. Moreover, a synthetic long-term dataset of CSI timeseries has been generated in [32] that is statistically indistinguishable from the observed data. The model of [32] requires an input from 10 to 15 annual time-series of hourly values that may be retrieved from satellite-based GHI data.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Moreover, a synthetic long-term dataset of CSI timeseries has been generated in [32] that is statistically indistinguishable from the observed data. The model of [32] requires an input from 10 to 15 annual time-series of hourly values that may be retrieved from satellite-based GHI data. The Hidden Markov Models (HMMs) with Gaussian observation have been used in [33] to generate synthetic CSI data.…”

Section: Literature Reviewmentioning

confidence: 99%

“…In addition, compared to [31], we develop an N-state MC model for each group of days with (i) cloudy, (ii) intermittent cloudy, and (iii) clear sky behavior, whereas a single N-state MC model has been developed in [31]. Compared to [32], where the synthetic data of PV systems power production have been used for analyzing the profitability/risk of financing PVs in long-term resolution by means of a model based on 10 to 15 years of real data, the model proposed in this paper is advantageous for numerical simulation in shortto mid-term analysis and it is trained by a limited amount of input measurements.…”

Section: Contributionmentioning

confidence: 99%

See 1 more Smart Citation

Generation Data of Synthetic High Frequency Solar Irradiance for Data-Driven Decision-Making in Electrical Distribution Grids

et al. 2021

View full text Add to dashboard Cite

In this paper, we introduce a model representing the key characteristics of high frequency variations of solar irradiance and photovoltaic (PV) power production based on Clear Sky Index (CSI) data. The model is suitable for data-driven decision-making in electrical distribution grids, e.g., descriptive/predictive analyses, optimization, and numerical simulation. We concentrate on solar irradiance data since the power production of a PV system strongly correlates with solar irradiance at the site location. The solar irradiance is not constant due to the Earth’s orbit and irradiance absorption/scattering from the clouds. To simulate the operation of a PV system with one-minute resolution for a specific coordinate, we have to use a model based on the CSI of the solar irradiance data, capturing the uncertainties caused by cloud movements. The proposed model is based on clustering the days of each year into groups of days, e.g., (i) cloudy, (ii) intermittent cloudy, and (iii) clear sky. The CSI data of each group are divided into bins of magnitudes and the transition probabilities among the bins are identified to deliver a Markov Chain (MC) model to track the intraday weather condition variations. The proposed model is tested on the measurements of two PV systems located at two different climatic regions: (a) Yverdon-les-Bains, Switzerland; and (b) Oahu, Hawaii, USA. The model is compared with a previously published N-state MC model and the performance of the proposed model is elaborated.

show abstract

“…These probabilistic methods allow modelling the changes and evaluate likelihoods of the simulated metric's probability of lying above or below a particular value or within a specific range [11].In this scenario, it takes importance the generation of synthetic solar datasets in order to extend the information related to the variability of the solar radiation. The advantage of the synthetic data series is that allow to represent the behavior of solar radiation in a great range of scenarios and facilitates the risk evaluation of solar projects [12].…”

Section: Introductionmentioning

confidence: 99%

Synthetic generation of plausible solar years for long-term forecasting of solar radiation

Jiménez-Valero

Larrañeta

López-García

et al. 2022

Theor Appl Climatol

View full text Add to dashboard Cite

En este trabajo se mejora y se valida un algoritmo propuesto en estudios previos para generar series sintéticas de años solares plausibles (PSY). La metodología proporciona 100 años sintéticos en paso minutal de radiación global horizontal (GHI) y radiación directa normal (DNI) acoplados. El algoritmo requiere un conjunto de datos de 10 a 20 años de GHI y DNI acoplados en paso horario que se pueden obtener para la mayoría de las ubicaciones del mundo a partir de estimaciones de satélite. El algoritmo se evalúa en cinco emplazamientos seleccionados por su diferente tipo de clima según la clasificación de Koppen-Geiger. La evaluación se realiza en diferentes escalas temporales: anual, mensual, diaria y minutal. En todos los casos, las series sintéticas de años solares plausibles cubren una gama más amplia de escenarios que las series observadas, pero manteniendo su distribución. Los resultados sugieren que los PSY generados sintéticamente son capaces de reproducir la variabilidad natural del recurso solar en cualquier emplazamiento facilitando la simulación estocástica de los sistemas de aprovechamiento solar.

show abstract

Generation of synthetic solar datasets for risk analysis

Cited by 16 publications

References 28 publications

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications: Third Edition

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications: Third Edition

Generation Data of Synthetic High Frequency Solar Irradiance for Data-Driven Decision-Making in Electrical Distribution Grids

Synthetic generation of plausible solar years for long-term forecasting of solar radiation

Contact Info

Product

Resources

About