Participation of battery storage systems in the automatic frequency restoration reserve market based on machine learning
Merten, Michael; Sauer, Dirk Uwe (Thesis advisor); Rehtanz, Christian (Thesis advisor)
Aachen : RWTH Aachen University (2020, 2021)
Book, Dissertation / PhD Thesis
In: Aachener Beiträge des ISEA 150
Page(s)/Article-Nr.: 1 Online-Ressource (xi, 187 Seiten) : Illustrationen, Diagramme
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020
Automatic Frequency Restoration Reserve (aFRR) is one of three control reserve services in continental Europe to compensate imbalances in the electrical grid. Due to the transition towards a high degree of renewable sources, fewer conventional power plants will be available in the future to stabilize the electrical grid. However, new technologies are facing various entrance barriers because of the complex design of the aFRR market. This thesis provides a detailed aFRR market analysis and derives revenue potentials for different bidding strategies. By not relying on conventional power plants to stabilize the electrical grid, this work contributes in enabling a higher share of renewable sources. To assist the creation of bidding strategies, a market prediction methodology is presented. For any potential bid, the acceptance probability in the next auction process is derived. Both statistical and machine learning based models are used for predicting key market quantities. An in-depth model comparison on numerous time series predictions reveals a usually better performance of statistical models. In only a few cases, the (recurrent) neural network models slightly outperform the statistical models. Exogenous data sources such as weather, electrical loads or market data did not significantly improve the prediction performance. This work further developed an operating strategy for integrating Battery Energy Storage Systems (BESS) into existing virtual power plants (VPP) to jointly participate in the aFRR market. The operating strategy exploits the advantages of multiple technologies by selling a maximum of the generation capabilities on the market while the BESS instantly responds to aFRR requests and can recharge from the generation units. Based on an optimization process and the results from above mentioned predictions, a bidding strategy is presented that optimizes the bids to submit to the aFRR auction. An in-depth cost breakdown and battery-ageing model support the derivation of optimal bids and earning potentials. Special focus is on VPPs comprising only renewable sources and a BESS. With current costs of containerized BESS, an operation is not economically viable. Compared to the provision of Frequency Control Reserve (FCR), profits on the aFRR market were found to be lower. However, with a predicted cost breakdown for the year 2025, the pooled operation can generate profits and can contribute to grid stabilization in times of high levels of renewable sources.