No abstract

Driven by the transformation of the energy system and the need to store fluctuating power generation from renewable energy sources, stationary storage systems become more and more important. [1] Flow batteries are promising candidates for balancing the demand and supply in the electrical grid and thus providing continuous and reliable power supply. The unique concept of flow batteries is based on the spatial separation of electrolyte and electrode which results in the independent scalability of power and capacity. [2,3] This makes them applicable in a wide energy range of the grid infrastructure. Moreover, long operating life cycles, environmental friendliness, and nonflammability are further advantages toward predominant storage systems. [4,5] However, high capital costs and low energy density predominantly have been preventing a deep market penetration up to the present. [6] A typical flow battery consists of two independent reservoirs holding separated electrolyte solutions and two porous electrodes separated by an ion transport membrane. During operation, the electrolytes are pumped through the electrochemical cell, where the redox reaction takes place at the surface of the porous electrodes. Subsequently, the charged or discharged electrolyte flows back into the reservoir. [7,8] The all-vanadium system is so far the most studied and developed flow battery system, which convinces by the absence of material degradation and capacity fade. [9,10] However, it suffers from the comparatively high prices and enormous price fluctuations regarding the electrolyte, [11] and from the fact that vanadium is a critical raw material. [12] Recently, organic redox-active materials have emerged as a promising alternative to metal-based systems due to their low cost, structural designability, and the independence on metal mining. [13,14] For this reason, this contribution focuses on an electrolyte for aqueous organic redox flow batteries, namely 4-hydroxy-2,2,6,6-tetramethylpiperidin-1-oxyl (4-OH-TEMPO). [11,15] Simulation methods are powerful tools for gaining insight into the multiphysical processes inside the battery and for predicting its performance for various operating conditions. They provide a substantial alternative to experimental investigations as these can be very challenging and costly in terms of raw materials, physical resources, and time. Modeling approaches for flow battery components and performance apply to a wide range of length scales. This contribution focuses on a detailed microscale model (MSM) and an upscale connection to a homogenized cell-scale model (HCSM).Three-dimensional MSMs resolve the actual electrode geometry to gain insights of the micro-structural processes within the battery. The structure of the electrode is obtained by image processing and reconstructing, mostly using X-ray-computed

Due to the characteristics of flow batteries, this technology is ideally suited for low-cost storage in the range of a few hours and thus for load balancing as stationary storage in grids with high amounts of renewable energy [1]. Today, a large number of different active materials for flow batteries are known, although only a few have been commercialised [2]. Basically, the energy supply and thus also the required storage should be sustainable, i.e. not cause resource problems and not be harmful to humans and the environment. A potential for a huge range of possibilities is offered by organic active materials, which should be used especially in aqueous solutions. Due to the immense possibilities, classical synthesis and testing is extremely lengthy and costly. An alternative can be model-based high-throughput screening, where by simulating the properties of active materials in the electrolyte and the battery itself, computer-based simulations can be used to conduct the search. The SONAR project is an EU-funded project in which 7 different institutions from the EU, Switzerland and Australia are developing a high-throughput screening method capable of finding new active materials for redox flow batteries. The principle is a serial coupling of different size scales, combined with molecule generation and machine learning. The chemical structure of a candidate is generated by a molecule generator and then its atomistic properties, kinetics, side reactions and cell properties are iteratively calculated with exclusion criteria. In this talk we will give an overview of 2 years of research in this project in the areas of machine learning for high throughput screening, DFT based quantum mechanics modelling, kinetics Monte Carlo methods for meso-scale, 0D cell modelling, 3D cell modelling, stack modelling and techno-economics. [1] B. Dunn, H. Kamath, J.-M. Tarascon, Science 2011, 334, 928–935. [2] J. Noack, N. Roznyatovskaya, T. Herr, P. Fischer, Angew. Chem. Int. Ed. 2015, 54, 9776–9809. Figure 1