Development of a methodology for evaluation of impact of interannual and long-term variabilities of solar resources on the analysis of financial risk of a photovoltaic solar plants

In a context of energy transition from carbon energy to carbon-free energy, the installation of solar power plants is strategic. This kind of installation represents a great investment which is mostly made at the start of the installation. Therefore, a financial risk assessment is realised. We have identified, among all sources of uncertainty related to this analysis, a necessity to take the long term variations of the irradiance into account. A statistical analysis and a literature review have shown that current methods used to characterize the irradiance on the lifetime of the installation are limited. In particular, the use of statistical indicators (e.g. P90) have been discussed. The hypothesis of temporal stationarity of the irradiance, which is assumed in most of the study, has been questioned. This led us to deepen the analysis of the long term variations of the irradiance. These variations have been characterized with the use of a time-frequency decomposition tool, that was developed during this PhD. This brought us to distinguish three classes of variability: the intra-annual variability, the annual to decadal variability, and the multi decadal variability. These three classes have been analysed for long term databases. The databases are of various kinds: long term measurements of the GEBA network, CLARA-A2 satellite data, MERRA-2 reanalysis data, and data from climate models IPSL-CM6A-LR. For this set of databases, variability has shown a latitudinal repartition: the inter-annual variability has a weak influence on the extra tropical zone, and a strong influence close to the equator. For the extra tropical zone, which is mostly influenced by the intra-annual variability, the use of four years historical data will be sufficient in order to take into account all the variation of the irradiance in a correct way. On the other hand, for the tropical zone influenced by the annual to decadal variability, the use of 30 years historical data is recommended. An alternative to the use of this great amount of data has been addressed, through a possible modelization of the characteristic oscillations of each scales. Regarding the multi decadal variability, the use of more than 30 years of data is encouraged. A large diversity of the variability structures has been observed, depending on the databases. In particular, the GEBA database shows a greater influence of the inter annual variability than the other databases. This result questions the representativity of the inter annual variations in the gridded databases. At this point of the research work, none of the gridded databases has stood out. For one location of interest, it is therefore recommended to consider them all, and to study the full range of variation scales. (author)