Prediction of atmospheric variables for solar energy management in West Africa

• Project Background

The Economic Community of West African States (ECOWAS) has identified energy transition, water security and sustainable agriculture as the three main priorities for meeting the commitments made by West African States in their Nationally Determined Contributions (NDCs) to the Paris Agreements. The energy strategy is to accelerate access to electricity for hundreds of millions of Africans by limiting the use of fossil fuels. Renewable energy technologies are therefore taking an increasing place in energy and climate policies in compliance with the Paris agreements. By 2030, while the electricity consumption of ECOWAS is expected to quadruple, the installed solar photovoltaic capacities will reach 8 to 20 GWp according to different scenarios (IRENA, 2018).
Despite one of the highest solar potentials in this region of the world (e.g. Broesamle et al., 2001 ; Diabate et al., 2004 ; Quansah et al., 2016), solar electricity production remains variable as it depends on meteorological conditions, making its integration complex with respect to the stability of electrical networks. Photovoltaic production is directly related to the solar irradiance received at the earth’s surface (the solar power per square meter reaching the ground), which is itself strongly influenced by the atmospheric dust load and the presence of clouds. In some regions of West Africa, the attenuation of solar radiation by clouds can be close to 50% during the monsoon (Danso et al, 2020). In addition, solar radiation there is subject to large spatio-temporal fluctuations in connection with meteorological phenomena of various scales such as mesoscale convective systems (MSC) and clouds at different altitudes (Schuster et al. 2006, Lafore et al. 2011, Bouniol et al. 2012).
Predicting the variability of the solar resource as accurately as possible on an intraday and daily scale is therefore essential to ensure optimal management of photovoltaic production by grid operators and power plants. While such forecasts are readily available in the most technologically advanced countries, they are still a particular challenge for West Africa. Available observations are sparse (especially in situ observations such as radio soundings), and weather forecasting systems are not necessarily developed. Few, if any, regional weather models exist and run operationally in this region. Several studies have still assessed the suitability of these models but either had low spatial resolution (Gbode et al., 2019a, 2019b) or focused only on short periods and specific events (Igri et al., 2018). The large variability of the scales involved and the importance of local phenomena (just a few hours and a few kilometers in the case of convective systems) classically involve the implementation of high-resolution meteorological models equipped with data assimilation schemes (Mathiesen et al. 2013, Kurzrock et al. 2019, Jimenez et al. 2022), with the ability to simulate cloud-aerosol-radiation interactions. Such systems are relatively expensive in computational time, euros, expertise, and carbon. They usually require the use of supercomputers that are not accessible to SMEs (Small and medium-sized enterprises) or specialized organizations in West Africa.
From another perspective, artificial intelligence (AI) has been developing exponentially over the past decade, enabling new applications. The major weather forecasting centers are investing massively in the exploration of these techniques for the outsourcing of numerous tasks in the forecasting chains (Duben et al, 2022). AI techniques are attractive because of their potential to reduce computation time in operational settings, but also because of their ability to perform forecasts in contexts that are very poorly formalized mathematically (dog/cat distinction, cloud recognition, etc.). However, due to the physical complexity of atmospheric processes and the need to have a large database for the calibration of these AI methods, they have not yet fully replaced the more conventional systems based on physical models. Thus, it seems premature to consider in the short term a weather forecasting system over West Africa based solely on AI.

• Objective of the project

The objective of the PhD thesis is to develop a solar resource forecasting system at the locations of several solar farms located in West Africa, taking into account all the constraints mentioned above : numerical sobriety is at the core of the scientific problem. High performance computing is not excluded for the experiments, but should not be the keystone of the finalized system since it will be integrated into an operational forecasting chain. In addition to the primary forecasts of solar irradiance, we will also be able to focus on the secondary variables impacting photovoltaic production (temperature, wind, humidity) and those that can modify the performance of the power plants via the soiling or cleaning of the panels (dust, wind, precipitation). The first phase of the work will consist in making a catalog of the existing data relevant for solar irradiance forecasting in West Africa : satellite observations and products (cloud types, optical thicknesses, estimated solar radiation), meteorological stations, etc. Depending on the nature and density of these data, the development of "non-physical" models (i.e. neural networks) could be attempted, which would constitute a strong originality of the thesis. Alternatively or subsequently, the development will continue with the implementation of the limited area model WRF (Weather Research and Forecasting) in its specific "solar" version (WRF-solar, Jimenez et al, 2016), over a region of West Africa. A data assimilation method backed by WRF will be implemented and calibrated for the region. AI-based techniques would then come into play either to reduce computation time (e.g., through super-resolution techniques), or for the transformation of complex variables (e.g., cloud cover) to facilitate the interface with the physical model. The replicability of the solution to other regions of the world will require particular attention.

• Scientific and programmatic framework

This thesis project is part of the NETWAT project (West African Mineral Dust : a key in the NExus climaTe - WATer -energy), financed by the ANR (Agence Nationale de la Recherche) and led by Sandrine ANQUETIN of the IGE. The thesis is funded by the project.

The thesis is a partnership between the IGE and the company Steadysun based in Bourget-du-Lac. The IGE is one of the largest research laboratories on terrestrial fluid envelopes in France.
Steadysun was created in 2013 by the French Atomic Energy and Alternative Energies Commission (CEA) and the French National Solar Energy Institute (INES). As a player in the international energy transition, Steadysun offers solar production forecasting services that help increase the share of solar energy in the electricity mix, while mitigating the costs and risks associated with weather variability. The value provided by these services addresses the needs of balancing grids, minimizing investment and operating costs of power plants, optimizing transactions on electricity markets and maximizing self-consumption. To date, the forecasting solutions proposed by Steadysun are deployed on more than 14000 sites in more than 25 countries, including 4 in West Africa (Senegal, Burkina Faso, Benin, Niger).

The thesis will be supervised by Emmanuel COSME (thesis director), associate professor at IGE and specialist in data inversion (assimilation, neural networks), Damien RAYNAUD (Steadysun co-supervisor), meteorologist, PhD in climate variability and integration of renewable energies, R&D engineer for weather forecasting systems at Steadysun.
Strong interactions will take place with some key people of the NETWAT project : Sandrine ANQUETIN at IGE, and Guillaume TREMOY at Steadysun (Research and Innovation Director).

One or more collaborative missions in West Africa may take place depending on the project developments.

• Expected profile

The candidate will have a background in scientific computing, or atmospheric and climate sciences, or data analysis. Whatever the initial training, experience in handling large datasets or in implementing a "heavy" model would be a definite asset. If in doubt about the suitability of the profile, the hesitant candidate is invited to contact the supervisors informally for advice. Excellent writing and communication skills in French and English are expected.

• How to apply

Send CV and cover letter to
emmanuel.Cosme(at)univ-grenoble-alpes.fr and damien.raynaud(at)steady-sun.com.

The CV should be specific about the experience related to the profile sought (above) and include the names of at least two references (internship supervisors, training supervisors).

The CV and letter must be received by May 15, 2023, 9:00 am. Pre-selected candidates will be interviewed between May 22 and June 2, 2023.

• Useful links and bibliographic references

IGE : www.ige-grenoble.fr
Steadysun : https://www.steady-sun.com/fr/
NETWAT project : https://netwat.osug.fr/

Bouniol, D., Couvreux, F., Kamsu-Tamo, P., Leplay, M., Guichard, F., Favot, F., & O’Connor, E. J. (2012). Diurnal and Seasonal Cycles of Cloud Occurrences, Types, and Radiative Impact over West Africa, Journal of Applied Meteorology and Climatology, 51(3), 534-553.
Broesamle, H. ; Mannstein, H. ; Schillings, C. & Trieb, F. Assessment of solar electricity potentials in North Africa based on satellite data and a geographic information system, Solar Energy, Elsevier BV, 2001, 70, 1-12
Danso, D. K. ; Anquetin, S. ; Diedhiou, A. & Adamou, R. Cloudiness Information Services for Solar Energy Management in West Africa, Atmosphere, MDPI AG, 2020, 11, 857
Diabaté, L. ; Blanc, P. & Wald, L. Solar radiation climate in Africa. Solar Energy, Elsevier BV, 2004, 76, 733-744
Duben et al, Machine learning at ECMWF : A roadmap for the next 10 years, ECMWF technical Memo 878, 2022. https://www.ecmwf.int/en/elibrary/81207-machine-learning-ecmwf-roadmap-next-10-years
Gbode, I. E., Ogunjobi, K. O., Dudhia, J., & Ajayi, V. O. (2019). Simulation of wet and dry West African monsoon rainfall seasons using the Weather Research and Forecasting model. Theoretical and Applied Climatology, 138, 1679-1694.
Gbode, I. E., Dudhia, J., Ogunjobi, K. O., & Ajayi, V. O. (2019). Sensitivity of different physics schemes in the WRF model during a West African monsoon regime. Theoretical and Applied Climatology, 136, 733-751.
Igri, P. M., Tanessong, R. S., Vondou, D. A., Panda, J., Garba, A., Mkankam, F. K., & Kamga, A. (2018). Assessing the performance of WRF model in predicting high-impact weather conditions over Central and Western Africa : An ensemble-based approach. Natural Hazards, 93, 1565-1587.
IRENA (2018), IRENA Planning and prospects for renewable power : West Africa, International Renewable Energy Agency, Abu Dhabi.
Jimenez, P. A., Hacker, J. P., Dudhia, J., Haupt, S. E., Ruiz-Arias, J. A., Gueymard, C. A., Thompson, G., Eidhammer, T., & Deng, A. (2016). WRF-Solar : Description and Clear-Sky Assessment of an Augmented NWP Model for Solar Power Prediction, Bulletin of the American Meteorological Society, 97(7), 1249-1264. doi : https://doi.org/10.1175/BAMS-D-14-00279.1
Jiménez, P. A., Dudhia, J., Thompson, G., Lee, J. A., Brummet, T., Improving the cloud initialization in WRF-Solar with enhanced short-range forecasting functionality : The MAD-WRF model, Solar Energy, Volume 239, 2022, Pages 221-233, ISSN 0038-092X, https://doi.org/10.1016/j.solener.2022.04.055.
Kurzrock, F., Assimilation de données satellitaires géostationnaires dans des modèles atmosphériques à aire limitée pour la prévision du rayonnement solaire en région tropicale. Géographie. Université de la Réunion, 2019. Français.
Lafore, J. P., Flamant, C., Guichard, F., Parker, D. J., Bouniol, D., Fink, A. H., ... & Thorncroft, C. (2011). Progress in understanding of weather systems in West Africa. Atmospheric Science Letters, 12(1), 7-12.Mathiesen, P., Collier, C. and Kleissl, J., A high-resolution, cloud-assimilating numerical weather prediction model for solar irradiance forecasting, Solar Energy, Volume 92, 2013, Pages 47-61, ISSN 0038-092X, https://doi.org/10.1016/j.solener.2013.02.018.
Mathiesen, P., Collier, C., Kleissl, J. (2013). A high-resolution, cloud-assimilating numerical weather prediction model for solar irradiance forecasting. Solar Energy. 92. 47–61. 10.1016/j.solener.2013.02.018.
Quansah, D. A. ; Adaramola, M. S. & Mensah, L. D. Solar Photovoltaics in Sub-Saharan Africa – Addressing Barriers, Unlocking Potential. Energy Procedia, Elsevier BV, 2016, 106, 97-110
Schuster, R., Fink, A. H., & Knippertz, P. (2013). Formation and maintenance of nocturnal low-level stratus over the southern West African monsoon region during AMMA 2006. Journal of the Atmospheric Sciences, 70(8), 2337-2355.

Updated on 17 mars 2023