Planification de l'électrification rurale décentralisée en Afrique subsaharienne à l'aide de sources renouvelables d'énergie : le cas de l'énergie photovoltaïque en République de Djibouti

摘要

Over the past 40 years, growth of renewable energies benefited of the new world energy frame, which resulted of the questioning about what development of human societies had to be. Furthermore, although human development comes with electricity, the rural condition of many populations of Sub-Saharan Africa incites us to look for suitable power supply alternatives. Eventually, in this specific context, renewable energies can represent a reliable solution to the off-grid electrification of rural people. However, this solution has to be economical and technical, and not only political. The Republic of Djibouti is a little developing country located in the Horn of Africa which perfectly symbolizes the social and energy challenges of rural populations in Sub-Saharan Africa. Instability and limitation of the existing electrical grid, fuel cost and lack of fossil resources point to the geographically diffused solar resource as probably the best way to improve human development and reduce poverty of Djiboutian rural peoples. Therefore, we have considered the study of photovoltaic (PV) systems within the rural off-grid electrification frame. Essentially, within this work, Djibouti was the study case of an overall scientific methodology, whose primary objective is to be reusable by countries showing similar energy and social characteristics. Firstly, in order to evaluate relevance of these systems, it was necessary to estimate the level and repartition of the solar resource across the country. So we developed a solar atlas, i.e. cartography of the hourly solar irradiation reaching the ground, based on satellite-derived irradiance estimates retrieved between 2008 and 2011. For assessing the atlas quality, we compared irradiation estimates against ground measures retrieved on 4 different sites by 2 temporary weather stations deployed between 2010 and 2013. This comparison globally showed good results with, for the daily case, a maximum relative root mean squared error of 8.05 % and a minimum correlation coefficient of 0.8892 through all samples. Finally, yearly map extracted from the atlas showed that, with a daily mean irradiation of 5.87 kWh/m^2.day, the solar potential of Djibouti is one of the most significant in the world. Satellite models are useful for determining solar irradiance at ground level but they don't take into account local topography effects. In order to incorporate these shading effects to the satellite-derived irradiance maps, improving therefore irradiance accuracy and spatial resolution, we used a Digital Elevation Model (DEM). Firstly, this disaggregation process was based on the development of a new fast horizon algorithm which was assessed by means of topographic measures in Corsica Island. This model thus shows higher performances than already existing ones, with not only a better precision but also a substantially reduced CPU time consumption, which makes this one a well suited model for calculation of large mapping areas. Finally, by correcting irradiance with horizon shading and elevation gradient for each pixel of the DEM, we improved geographic information of the solar irradiation atlas. Although solar resource is the first indicator of the photovoltaic potential, other elements, as environmental parameters or endogenous characteristics of photovoltaic modules, also have to be taken into consideration to precisely estimate the energy produced by a PV system. Hence, by means of different models, we evaluated the influence of irradiance and temperature onto the conversion efficiency of a PV generator to finally retrieve the atlas of the PV potential across the country. Furthermore, we developed an empirical method to retrieve ambient temperature from land surface temperature (LST), by training an artificial neural network with satellite estimates and ground measurements. In conclusion, by combining this cartography to a multi-criteria approach comparing relevance between PV systems and classical power supply systems within the rural electrification scheme, we developed the first photovoltaic decision making tool of the country intended for all officials acting in the energy field.