Probabilistic modelling of renewable generation to account for uncertainties in interconnection studies

The development of renewable generation connected to the distribution network can create overvoltage or overcurrent situations leading to expensive upgrades of the network. Solutions such as production curtailment could avoid these expenditures, but in order to take full advantage of these new solutions, one must first appropriately model the statistical behaviour of loads and generation over several years with a high level of uncertainties. This paper performs a statistical analysis on 4254 real-world generation time series that run over about 2 years, with a 10-minute time step, for four different types of generation technologies. This large dataset was made available thanks to a partnership with the major French distribution system operator (DSO), Enedis. The objective is to develop probabilistic models of the power output for each type of generator and also of the statistical dependency between two generators and between generation and load and to apply them on an interconnection case study.