Stage Conception d’un système de prévision de l’énergie solaire h/f

Design of a solar energy forecast system based on an optimal blending of different source
Préambule :
The clean energy transition must leverage on an energetic mix, largely supplied by distributed generation (DG) based on variable renewable energies (VRE), such as solar and wind energy. It results that the electricity systems will face exceptional challenges in the coming years.
One of these challenges is to achieve the energy transition targets whilst continuously meeting the end-user demand at acceptable prices. Since VRE are intrinsically variable, uncertain and non-dispatchable, the stakeholders of the european energy system (transmission and distribution system operators, TSOs and DSOs, utilities, small and large energy suppliers, etc.) have to choose adequately the type and size of VRE installations and associated storage systems.
It will also be important to forecast precisely the amount of energy that could be produced in the future to optimally manage the flexible power production assets. In this context isolated areas, such as Reunion Island, which already host a significant share of VRE (25% of the installed capacity), are pioneers and they are ideal to be considered as living labs to develop and test innovative solutions, such as a smart solar microgrid, in order to accelerate the energy transition and supporting Europe in facing its future energy problems.

Context :
The project TwInSolar is funded by the european commission horizon Europe research and innovation programme. TwInSolar aims at strengthening the research and innovation capacities of the UR that belongs to the outermost region (OR) of Europe in order to create new opportunities inside and outside of Europe and more specifically in the zone surrounding Reunion.
This capacity building will be done through the establishment of effective partnerships with the technical university of Denmark (DTU) and the fraunhofer institute for solar energy systems (Fraunhofer-ISE), which are internationally renowned research organizations, and with the support of the regional R&I agency of Reunion Island nexa and the conference of peripheral maritime regions (CPMR). Particular emphasis will be placed on issues related to the massive integration of solar energy production in insular territories.
The TwInSolar project includes a research activity related to the design and the operation of smart microgrid situated in the Reunion Island. The aim of this activity is to design an innovative and affordable solution to supply energy to the university campus of Terre-Sainte, Reunion (see figure and table below) with a massive integration of solar power generation. With two net-zero energy buildings (enerpos and ESIROI) on this site, UR has already set up an efficient energy saving policy for this campus.
The expected results will go one step further by studying a solution to significantly reduce the carbon emissions by feeding the microgrid with at least 80% of on-site solar production. One of the main outcomes is the generation of short-term probabilistic forecasts required to carry out a predictive control of the micro-grid. During this research activity, Fraunhofer-ISE and UR will be involved in the co-supervision of the intern hired by UR.
For generating solar power forecasts, it has been demonstrated that an optimized forecast can be obtained for each horizon by optimally blending different sources of input such as irradiance ground measurements, satellite data and numerical weather predictions.

Purpose of work :
Fraunhofer will install on-site a sky-imager to generate very short-term forecasts (several second up to 30 minutes). UR and Fraunhofer will develop and benchmark a statistical forecasting model that will optimally blend the solar forecasts generated by sky-imager with past ground data, satellite forecasts and numerical weather predictions. The aim will be to get rapid updates of probabilistic forecasts ranging from several seconds to 3 days ahead.

Expected activities :

• Collection of data from different sources (ground data, sky imager, satellite derived forecasts and NWP).
• Development of statistical model (machine learning) to optimally blend the different sources of data for each forecast horizon.
• Evaluation of the quality of the generated forecasts.

Profil :

• Vous êtes en Master (1ère ou deuxième année) ou en école d’ingénieurs (4ème ou 5ème année).
• Vous cherchez un stage dans le monde de la recherche dans le domaine de la transition énergétique avec une dimension internationale.

Duration : between 4 and 6 months.
Location : campus de Terre-Sainte, Saint-Pierre, La Réunion.