Stable, high-capacity flow batteries could power grid-scale renewable energy storage
Transitioning to renewable energy sources like solar and wind is becoming increasingly critical in the fight against CO2 levels. However, these sources are not always available – the sun does not always shine and the wind does not always blow. Substantial energy storage is crucial to ensure a consistent power supply. However, storing excess energy for use during periods of low production poses certain challenges. “The traditional method of energy storage – pumping water uphill and letting it run through a turbine to generate electricity – is not practical in many locations. Furthermore, lithium-ion batteries, while useful, are expensive and provide energy for only up to four hours,” notes Pekka Peljo, coordinator of the EU-funded CompBat project. Flow batteries, which store energy in liquid electrolytes rather than electrodes, offer a valuable alternative. “By simply increasing the volume of the tanks storing the liquid, the energy storage capacity can be increased,” adds Peljo.
Chasing a fast track to high-performance, low-cost active materials
Current flow batteries rely on costly active materials, which prompts the need for more affordable alternatives. “Our main objective was to seek low-cost active materials that meet often conflicting requirements of sufficient cell voltage, high solubility and long-term stability for around 20 years of operation,” remarks Peljo. Typically, experimental approaches are guided by researchers’ intuition, resulting in an initial selection of molecules for testing. The first steps involve the synthesis, purification and characterisation of these candidate molecules. Based on the results of the testing phase, the molecules may be modified and synthesised again in a continuous circle. However, this process is rather slow.
Artificial intelligence bridges the gap from intuition to innovation
CompBat has created data-driven approaches to accelerate material development. “We have developed machine learning methods for the quick screening of potential active materials for flow batteries and modelling tools for optimising cell and stack design,” explains Peljo. Project partners in Hungary developed an efficient computational protocol that could evaluate the redox potentials of 100 molecules per day using quantum chemical calculations performed on high-performance office computers. This process allows us to accurately evaluate the redox potential of new molecules – a property critical to determining battery voltage. After evaluating around 15 000 molecules, they compiled a substantial database of computational redox potentials. This database was then used to train different machine learning models capable of predicting the redox potential of virtually any organic molecule within seconds. “While these predictions still require experimental verification for selected molecules, the accuracy of the machine learning tool has been surprisingly good,” highlights Peljo.
The next wave of advances
This tool, coupled with cell-, stack- and system-level modelling tools, enhances understanding of the material requirements for commercially viable systems. Now molecules can be screened based on desired ranges of redox potentials and then selected for experimental testing. The tool is currently being used to design new materials, revealing how different modifications to the molecule structure affect the redox potentials. However, there are still challenges to overcome. Stability, for example, remains a concern. “Out of the 20 molecules synthesised throughout CompBat, only five demonstrated a certain level of stability, and only one or two could be used in a battery. Computational methods to evaluate stability are not as advanced as those for redox potentials, requiring further research,” concludes Peljo.
Keywords
CompBat, redox potential, flow batteries, machine learning, energy storage, active material, stability
