Modelling of the economic feasibility of largescale electricity storage technologies : a german case study

Abstract

Electricity storage is often portrayed as the solution for the challenges that an increasing capacity of intermittent generation brings. If the electricity storage facilities are to be introduced in the grid by private investors though, just like any other asset, they require a business case. This study is part of the DNV GL StRe@M project whose goals include the modelling of the economic feasibility of electricity storage facilities in future German electricity grid scenarios from a price-taking investor’s perspective by comparing costs and revenues. The two revenue streams considered in the StRe@M project come from the spot and reserve market, and this study focuses on modelling the latter for Germany. This thesis also provides a cost and revenue framework to assess the revenues from both markets and the resulting profits. First a qualitative study maps the German reserve market and the characteristics of its products to identify opportunities for electricity storage and the impacts of regulation thereon. Next a quantitative model is designed to assess the revenue potential of the future secondary reserve market by forecasting its demand and price levels. The modelling scope is limited to the secondary reserve (energy) market (named aFRR in Germany) only because of its relative market size, the low number of participants and data availability. A bottom-up approach was tried by looking for a quantified relation between (1) historical time series of forecast errors for load and solar and wind generation and (2) system imbalances or activated aFRR directly – a positive causal relation which often appears in literature. As no quantified relation could be found an alternative top-down stochastic approach then used the historical probability distribution of activated aFRR in 2015 to establish a stochastic function for aFRR demand in future scenarios up to a few years, preserving the properties of the historical probability distribution. An effort was made to scale this stochastic function for an increasing renewable penetration but no workable scaling could be obtained. The future prices to accompany the forecasted volumes were determined from a regression analysis on historical aFRR price time series. Regression components included the aFRR volume and the spot price. The design of the cost and revenue framework, used to process the potential revenues from the spot and reserve market, was based on comparing samples of a stochastic reserve market revenue with a deterministic spot market revenue and aggregating this into a distribution for the profit. To conclude the first dispatch and profit results of the StRe@M modelling are presented for a German electricity scenario in February 2020 with an 80% RES share, which should be used with great caution. The modelled lithium-ion battery technology and variablespeed PSH show positive profits on average, but the fixed-speed PSH does not. The main limitation of this model is the lack of the scaling effect for renewable penetration, for which a scenario analysis is probably most suited.