Long-term modelling of electricity market prices to examine prospective revenues of storage agents

Long-term modelling of electricity market prices remains a challenging task. Fundamental models can readily account for the drivers of mean electricity price level movements, but often fail to capture essential short-term price curve characteristics like high volatility and negative prices. Without such level of detail for the modelled prices, conclusions about the profitability of storage operations from an actor's perspective are hard to obtain. Here, we present a hybrid fundamental-econometric approach capable of reproducing both the short-term stylized facts and the long-term overall levels of day-ahead wholesale market prices obtained at the EEX with a high degree of accuracy, including a low mean average error, the reproduction of negative prices and a high volatility. The model considers the bidding curves of the technology classes nuclear, lignite, hard coal, gas combined cycle, gas turbine, oil, solar, wind, biomass, run-of-river hydro and pumped storage in a fundamental way. Plant capacities, efficiencies, feed-in profiles and load series for Germany are derived from freely available data sources. Using a genetic algorithm, we estimate ranges for bidding markups of the above mentioned power plant classes. Markups are allowed to be negative as well. Reasons are e.g. ramping costs, constraints from other markets (thermal, reserve markets) or feed-in tariff remuneration. The genetic algorithm follows multiple fitness criteria to adjust modelled electricity prices towards the calibration data set. These fitness criteria include the correlation and mean average error, as well as the similarity of standard deviation, minimum, mean and maximum of price values and a similar number of hours of negative prices. The calibration is performed with price data for the years 2012 to 2014. The validation is based on price data for the years 2015 and 2016. Descriptive statistics show a very high degree of accuracy for most shape parameters of the price curve (see sample results). This bidding mechanism has been implemented into an agent-based model to study the integration of renewable energy sources into the market. In this contribution, we want to assess the success of non-cooperative storage operators at the future electricity market with rising shares of variable renewable energy sources and increasing storage capacities. Questions in this context include: What is the impact of the modelled markups on the revenues of storage agents? How do available storage capacities affect the modelled electricity prices