Cette thèse s’inscrit dans le cadre d’un projet européen visant à améliorer la compétitivité des centrales solaires à concentration. Parmi les différents défis soulevés par ce projet, l’évaluation en temps réel de la disponibilité et de la variabilité de la ressource solaire est un point clé puisqu’elle permettrait une gestion optimisée du champ solaire et, par conséquent, une hausse de la productivité de la centrale. L’objectif de ce travail est donc de développer un outil d’évaluation de la ressource solaire destiné à la gestion de centrales CSP. Pour y parvenir, une étude approfondie des interactions entre le rayonnement solaire et l’atmosphère est tout d’abord menée. Cette étude révèle entre autres que l’éclairement normal direct (DNI) peut se scinder en deux composantes : le DNI par ciel clair et l’indice ciel clair. Le premier représente l’éclairement normal direct reçu au niveau du sol, lorsqu’aucun nuage ne vient occulter le Soleil. Le second traduit l’influence des nuages sur ce rayonnement par ciel clair. Évaluer ces deux composantes est essentiel pour l’opérateur de la centrale car elles lui permettent de connaître les marges de manoeuvre dont il dispose. D’une part, un modèle ciel clair permettant d’estimer et prévoir le DNI par ciel clair en temps réel est donc développé. Il permet de maintenir l’erreur quadratique moyenne sur l’estimation du DNI par ciel clair aux alentours de 30W/m². D’autre part, une caméra hémisphérique a été installée sur le site du laboratoire PROMES-CNRS afin de détecter les nuages et leur mouvement dans le but d’appréhender la variabilité de l’indice ciel clair. Ce système est notamment capable de fournir des images à haute dynamique, permettant de mesurer simultanément des informations dans la zone circumsolaire et dans les zones les plus sombres du ciel. Sur la base du modèle ciel clair et des images fournies par la caméra, un modèle de prévision du DNI pour tout type de conditions a été mis au point. Il permet de maintenir l’erreur quadratique moyenne sur la prévision du DNI aux alentours de 180 W/m², pour des horizons inférieurs à 30 min. En partenariat avec Acciona, l’outil développé est d’ores et déjà opérationnel sur la centrale solaire Palma del Rio II, en Espagne. / This thesis is part of a European research project which aims at improving the solar power plant efficiency. Among the different challenges pointed out by this project, the solar resource assessment and forecasting are essential tasks since they would allow a better real-time management of the solar field, and thus reduce the maintenance activities, while improving the expected benefits. Therefore, the purpose of this work is to develop a solar resource forecasting tool in order to improve the CSP plants management. An extensive review of the interactions between solar radiation and the atmosphere is firstly conducted. It reveals, among other things, that the direct normal irradiance (DNI) can be divided into two components : the clear sky DNI and the clear sky index. The former represents the direct normal irradiance received at ground level, when no clouds are occulting the sun. The latter reflects the influence of clouds on the clear sky DNI. Estimating these two quantities is essential for the plant operator, since it allows a better management of the solar field. As a consequence, a clear sky model able to estimate and forecast the clear sky DNI has been developed. The root mean squared error of the forecast is around 30 W/m². On the other hand, a sky imager has been installed at the PROMES-CNRS laboratory in order to detect clouds and their motion. The system is able to provide high dynamic range images, allowing the measurement of information both into the circumsolar area and into the darkest parts of the sky. Based on the clear sky model and the images provided by the sky imager, a DNI forecasting model is proposed. The root mean square error on the forecast is around 30 W/m², for 30 min forecasting horizon. One system is now operational at a solar power plant located in Palma del Rio II, Spain.

Global economic trends suggest that there is a need to generate sustainable renewable energy to meet growing global energy demands. Solar energy harnessed by concentrated photovoltaic (CPV) systems has a potential for strong contributions to future energy supplies. However, as a relatively new technology, there is still a need for considerable research into the relationship between the technology and the solar resource. Research into CPV systems was carried out at the University of Ottawa’s Solar Cells and Nanostructured Device Laboratory (SUNLAB), focusing on the acquisition and assessment of meteorological and local solar resource datasets as inputs to more complex system (cell) models for energy yield assessment. An algorithm aimed at estimating the spectral profile of direct normal irradiance (DNI) was created. The algorithm was designed to use easily sourced low resolution meteorological datasets, temporal band pass filter measurement and an atmospheric radiative transfer model to determine a location specific solar spectrum. Its core design involved the use of an optical depth parameterization algorithm based on a published objective regression algorithm. Initial results showed a spectral agreement that corresponds to 0.56% photo-current difference in a modeled CPV cell when compared to measured spectrum. The common procedures and datasets used for long term CPV energy yield assessment was investigated. The aim was to quantitatively de-convolute various factors, especially meteorological factors responsible for error bias in CPV energy yield evaluation. Over the time period from June 2011 to August 2012, the analysis found that neglecting spectral variations resulted in a ~2% overestimation of energy yields. It was shown that clouds have the dominant impact on CPV energy yields, at the 60% level.

Global economic trends suggest that there is a need to generate sustainable renewable energy to meet growing global energy demands. Solar energy harnessed by concentrated photovoltaic (CPV) systems has a potential for strong contributions to future energy supplies. However, as a relatively new technology, there is still a need for considerable research into the relationship between the technology and the solar resource. Research into CPV systems was carried out at the University of Ottawa’s Solar Cells and Nanostructured Device Laboratory (SUNLAB), focusing on the acquisition and assessment of meteorological and local solar resource datasets as inputs to more complex system (cell) models for energy yield assessment. An algorithm aimed at estimating the spectral profile of direct normal irradiance (DNI) was created. The algorithm was designed to use easily sourced low resolution meteorological datasets, temporal band pass filter measurement and an atmospheric radiative transfer model to determine a location specific solar spectrum. Its core design involved the use of an optical depth parameterization algorithm based on a published objective regression algorithm. Initial results showed a spectral agreement that corresponds to 0.56% photo-current difference in a modeled CPV cell when compared to measured spectrum. The common procedures and datasets used for long term CPV energy yield assessment was investigated. The aim was to quantitatively de-convolute various factors, especially meteorological factors responsible for error bias in CPV energy yield evaluation. Over the time period from June 2011 to August 2012, the analysis found that neglecting spectral variations resulted in a ~2% overestimation of energy yields. It was shown that clouds have the dominant impact on CPV energy yields, at the 60% level.

The description of a novel device to measure the spectral direct normal irradiance is presented. The solar spectral irradiance meter (SSIM) was designed at the University of Ottawa as a cost-effective alternative to a prohibitively expensive field spectroradiometer (FSR). The latter measures highly-varying and location-dependent solar spectrum, which is essential for accurate characterization of a concentrating photovoltaic system’s performance. The SSIM measures solar spectral irradiance in several narrow wavelength bands with a combination of photodiodes with integrated interference filters. This device performs spectral measurements at a fraction of the cost of a FSR, but additional post-processing is required to deduce the solar spectrum. The model was developed to take the SSIM’s inputs and reconstruct the solar spectrum in 280–4000 nm range. It resolves major atmospheric processes, such as air mass changes, Rayleigh scattering, aerosol extinction, ozone and water vapour absorptions. The SSIM was installed at the University of Ottawa’s CPV testing facility in September, 2013. The device gathered six months of data from October, 2013 to March, 2014. The mean difference between the SSIM and the Eppley pyrheliometer was within ±1.5% for cloudless periods in October, 2013. However, interference filter degradation and condensation negatively affected the performance of the SSIM. Future design changes will improve the longterm reliability of the next generation SSIMs.

Diploma thesis is concerned with business plan of creating production company solar resource. It specifies renewable resources of energy, business plan’s structure and analysis of external and internal environment of company in theoretical frame. Based on processing of critical analysis it proposes business plan of creating solar resource.