Standford scientists use AI to develop solid state lithium battery to get rid of battery explosion
Publish：Sino Energy Corporation Time：2017-01-02
[lithium engineering comprehensive report]
About lithium battery safety issues would have to mention so many consumers fear Galaxy Note 7 mobile phone battery explosion, Samsung to recall nearly 2 million sets of Note 7 mobile phone. This catastrophic event makes Samsung‘s third quarter net profit fell by 16.8%, as well as the mobile sector quarterly profits hit a record since the first launch of the Galaxy series of mobile phone lines since six years ago.
In order to improve the safety performance of lithium batteries, scientists in the past few decades have been looking for alternatives to lithium ion batteries in flammable electrolyte safety materials. Recently, the Stanford University researchers through the method of artificial intelligence and machine learning, found 20 kinds of solid electrolyte, allegedly they may replace the application in flammable electrolyte in the mobile phone, notebook and other electronic devices in the future, in order to improve the safety performance of the battery. The results of the study were published recently in the journal energy and environmental sciences.
The first author of the study, the leader of the study, the application of physics doctoral student Austin Sendek said the electrolyte in the battery played a role in the positive and negative transfer of lithium ions between electrodes. The electrolyte is low in cost and good in conductivity. However, once the battery is overheated or short circuited, the electrolyte is very easy to catch fire. Sendek believes that the most important advantage is the high stability of solid electrolyte, and compared to organic solvents, solid is not easy to explode and volatile, but also makes the structure of the battery is stronger.
In the process of long time past experiments and trial and error, scientists have failed to find and electrolyte as cheap, and has a good amount of solid electrolyte conductivity at room temperature. The Standford team did not use the traditional method of testing, random testing of individual compounds, but the use of artificial intelligence and machine learning.
The team builds a prediction model through experimental data. Based on the available data, the members have developed a computer algorithm to identify the quality of solid electrolyte compounds. This process is very similar to face recognition, the machine in the observation of several samples given, through in-depth learning to improve the accuracy of identification.
"There are tens of thousands of known lithium containing compounds," said Sendek. "We have only a small part of the test, but some of them may be the perfect conductor we are looking for. Therefore, we have established a computational model, which is based on the existing data we have learned, can be selected from the huge database of compounds to meet the requirements of our material. And this method is about 1 million times faster than our existing screening methods."
In order to build such a model, Sendek spent about two years to collect all of the scientific data related to the presence of solid state compounds containing lithium, a large amount of engineering can be imagined. Senior author of the paper, materials science and Engineering Assistant Professor Evan Reed said in an interview in order to establish the model of Austin, collected all the human wisdom and the related materials, and over the past few decades of measurement and experimental data.
It is these data that allow Austin to build a predictive model that predicts that a material will be a good solid electrolyte. This method allows us to screen all the lithium compounds at a relatively fast rate, so as to find the best candidate materials for further study".
This prediction model uses several criteria to screen candidate materials, including stability, cost, rich content, lithium ion conductivity and the ability to re route the electrons in the battery circuit. Candidate materials are selected from the "The Materials Project" database, and scientists can explore the physical and chemical properties of thousands of materials from the database.
Sendek told reporters, "we screened more than 12000 compounds containing lithium, singled out the potential of more than 21 solid electrolytes, the entire screening process takes only a few minutes. Most of my time is actually used to collect and manage all the data, and to develop metrics to define the accuracy of model predictions." The researchers decided to test the 21 materials in the laboratory to determine which material would be better suited to the actual situation.
"Our approach can solve a variety of material related issues and can enhance the effectiveness of research in these areas," Reed said. With the increase in the amount of data and computer technology, our ability to innovate will also be exponential form."
