https://hal-ifp.archives-ouvertes.fr/hal-03187020Martin de Lagarde, CyrilCyrilMartin de LagardeUniversité Paris Dauphine-PSL - PSL - Université Paris sciences et lettresENPC - École des Ponts ParisTechLEDa - Laboratoire d'Economie de Dauphine - IRD - Institut de Recherche pour le Développement - Université Paris Dauphine-PSL - PSL - Université Paris sciences et lettres - CNRS - Centre National de la Recherche ScientifiqueLantz, FrédéricFrédéricLantzIFP SchoolIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesImpact of Variable Renewable Production on Electriciy Prices in Germany : A Markov Switching ModelImpact of Variable Renewable Production on Electriciy Prices in Germany : A Markov Switching ModelHAL CCSD2017[SHS] Humanities and Social Sciences[SHS.ECO] Humanities and Social Sciences/Economics and FinanceBelli, Catherine2021-03-31 15:50:212023-03-24 14:53:212021-04-01 12:31:21enPreprints, Working Papers, ...application/pdf1This paper aims at assessing the impact of renewable energy sources (RES) production on electricity spot prices. To do so, we use a two-regime Markov Switching (MS) model, that enables to disentangle the so-called “merit-order effect” due to wind and solar photovoltaic productions (used in relative share of the electricity demand), depending on the price being high or low. We find that there are effectively two distinct price regimes that are put to light thanks to an inverse hyperbolic sinetransformation that allows to treat negative prices. We also show that these two regimes coincide quite well with two regimes for the electricity demand (load). Indeed, when demand is low, prices are low and the merit-order effect is lower than when prices are high, which is consistent with the fact that the inverse supply curve is convex (i.e. has increasing slope). To illustrate this, we computed the mean marginal effects of RES production and load. On average, an increase of 1GW of wind will decrease the price in regime 1 (resp. 2) by 0.77€/MWh (resp. 1€/MWh). The influence of solar isslightly weaker, as an extra gigawatt lowers the price of 0.73€/MWh in period 1, and 0.96€/MWh in regime 2. On the contrary, if the demand increases by 1GW in regime 1 (resp. 2), the price increases on average by 0.93€/MWh (resp. 1.18€/MWh). Although we made sure these marginal effects are significantly different from one another, they are much more variable than the estimated coefficients of the model. Also, note that these marginal effects are only valid inside each regime when there is no switching. The latter regime partly corresponds to the high load regime, at the exception of periods during which RES production is high. The impact on volatility could also be observed: the variance of the (transformed) price is higher during the high-p the switching of the coefficients, we allowed the probabilities of transition between the two regimes to vary over time, following a binomial logistic link with the relative share of RES production. This analysis shows that both wind and solar productions have a significant impact on the switching mechanism, especially on the probability of switching from the high-price regime to the low-price one, and consequently on the expected duration of each regime. However, the effect of wind pr on the probabilities is much higher than the effect of solar production, whereas they have a rather similar marginal effect on the price. Finally, although the regimes are sometimes highly correlated with some hours of the day, their endogenous determination (opposed to a semi-deterministic approach with dummy variables, for example) gives flexibility and keeps the model parsimonious.