École Nationale Supérieure de
Techniques Avancées

Abstract

A big challenge of sustainable power systems is to integrate climate variability into the operational and long term planning processes. Translating time series of meteorological data into time series of run-of-river based hydro power generation is a very delicate process, so we must decipher the complex relationship between the availability of water and the generation of electrical energy. In fact, we must take into account in this study several constraints. Among these constraints, we note on the one hand that the flow of water is a non-linear function of climate variables. On the other hand, physical phenomena such as melting snow act after a certain time on the runoff, and so it is clear that the impact of meteorological variables occurs with a certain delay. This work is part of the European project CLIM2POWER. It aims to translate climate data into energy systems and power indicators and then select the best techniques to provide power system operators with accurate data and operational guidance for power plants. Thus, this work aims to develop effective machine learning models for the prediction of run-of-river hydro power generation from climate variables. These models will be neural networks.