Assessing the probability of extremely rare renewable production and residual load in Europe at sub-seasonal to seasonal time scales

In the coming decades, the European energy system will undergo major transformations, like widespread electrification and fast development of variable renewable energy, in order to reach carbon neutrality and comply with the Paris Agreement. As a consequence, energy production will increasingly depend on weather variability, and the future European energy system needs to be designed to cushion this variability, at all time scales. In particular, rare events leading to extreme fluctuations in energy production or demand can be expected to play a major part in this design. For instance, one of the main challenging events is a combination of low renewable energy production with high demand for a long time period. To know how much flexibility will be needed (seasonal storage such as green gas reserves, number of auxiliary thermal power plants, needs in terms of demand-side management, etc.) and assess its cost, one needs to estimate the probability of occurrence of such events. However, observations of renewable energy production or climate variables are too short to quantitatively study these critical events. Therefore we need to rely on climate and energy models, and to develop dedicated tools to study extreme events of energy production and demand.

Here, we study the extreme imbalance between renewable generation and demand in Europe, at the sub-seasonal to seasonal scale. Based on a state-of-the-art climate model (CESM1.2.2) with extremely long simulations, we couple models of wind, solar PV, and demand with climate variables to obtain very long time series of energy production and demand. We consider 9 scenarios of renewable installed capacities to assess the probability of occurrence of extreme residual loads (demand minus renewable production). We study the statistics of extremely rare events that last for several weeks to several months.

The results show that extremely high residual loads are dominated by extremely low wind energy production events in winter, that are not visible in historical data. Leveraging our very long time series, we compute return time curves for extreme wind energy fluctuations. These curves tell us how frequent energy shortfalls of a given amplitude are. We find a renormalization such that return time curves depend weakly on the scenario. The estimation of such return times relies crucially on the available amount of data. We show that good approximations can be obtained from simple stochastic processes.