Energy News Monitoring

Designing a battery with optimal components can be a challenging task, as there are numerous ways to achieve a decent result. The most difficult part is to find the components that work best together. Even with the most modern computers, scientists so far haven’t been able to precisely model the chemical characteristics of every molecule that could prove to be the basis of a next-generation battery material.

Now (2019), researchers at the U.S. Department of Energy’s (DOE) Argonne National Laboratory have turned to the power of machine learning and artificial intelligence to speed up the process of discovering a new battery. They first created a highly accurate database of roughly 133,000 small organic molecules that could form the basis of battery electrolytes. To this effect, they used a computationally-intensive model called G4MP2. This collection of molecules, however, represented only a small subset of 166 billion larger molecules that scientists wanted to look at as electrolyte candidates. The research team also used a machine learning algorithm to relate the precisely known structures from the smaller data set to much more coarsely modelled structures from the larger data set.

To provide a basis for the machine learning model, the scientists used a modelling framework based on density functional theory, a quantum mechanical modelling framework used to calculate electronic structure in large systems. Density functional theory provides a good approximation of molecular properties but is less accurate than G4MP2.

Refining the algorithm to better ascertain information about the broader class of organic molecules involved comparing the atomic positions of the molecules computed with the highly accurate G4MP2 versus those analysed using only density functional theory.

Using AI to improve battery components and life is a relatively new discipline. In 2016, researchers at Stanford University used AI and machine learning to build predictive models for batteries from experimental data. They trained a computer algorithm to learn how to identify good and bad compounds based on existing data, much like a facial-recognition algorithm learns to identify faces after seeing several examples.

The scientists were convinced that the number of known lithium-containing compounds was in the tens of thousands, the vast majority of which were untested. They believed that some of them could be excellent conductors. Thus, they developed a computational model that learned from the limited data they already had, and then allowed them to screen potential candidates from a massive database of materials about a million times faster than current screening methods.To build the model, they spent more than two years gathering all known scientific data about solid compounds containing lithium.

In 2017, scientists carried out research on Battery Management Systems (BMS). They focused on modelling various types of batteries which, when implemented into the BMS, could give an insight into their performance. Parameters of three common models for various types of batteries were identified. Moreover, a common method that gives an insight into the lifespan of any battery under examination was found. This technique was based on several measurements taken at the laboratory and relied on using the Bayesian classifier for finding the state of health of a tested battery.

The advantages of this new algorithm are numerous: The machine learning algorithm provides a way to look at the relationship between the atoms in a large molecule and their neighbours and see how they bond and interact, and look for similarities between those molecules and others that have well-known properties.This will help make predictions about the energies of these larger molecules or the differences between the low- and high-accuracy calculations. By using G4MP2 as a gold standard the researchers could train the density functional theory model to incorporate a correction factor, improving its accuracy while keeping computational costs down.

This project was designed to give the biggest picture possible of battery electrolytes. ​If molecules for energy storage applications are used, it is important to know properties like their stability. This machine learning tool can be used to predict properties of larger molecules more accurately.