Using AI technology has become an importantway to develop new batteries. Foreign media reported that researchers atCambridge University and Newcastle University devised a new method to detectbatteries by sending electrical pulses to the batteries and measuring theresponse. The measurement results are processed by machine learning algorithmsto predict the health and service life of the battery.
The researchers claim that the technologycan predict the health of batteries with an accuracy that is 10 times higherthan current industry standards, thereby helping to develop safer and morereliable batteries for electric vehicles and consumer electronics.
It is difficult to predict the healthstatus and remaining service life of lithium-ion batteries is one of the mainproblems that limit the widespread adoption of electric vehicles. After runningfor a period of time, the lithium battery will decline, which affects thebattery's use state and life.
Current methods of predicting batteryhealth are based on tracking current and voltage during battery charging anddischarging, which misses an important function to indicate battery status.Therefore, there is currently a need to have new methods for tracking the manyprocesses that take place in the battery and to detect the actual operation ofthe battery, as well as new algorithms that can detect subtle signals duringcharging and discharging.
Dr. Alpha Lee of the Cavendish Laboratoryat Cambridge University said: "Safety and reliability are the mostimportant design standards because the batteries we develop can pack a lot ofenergy in a small space. By improving the monitoring of charging and dischargingSoftware, and using data-driven software to control the charging process, Ibelieve we can greatly improve battery performance. "The researchersdevised a method to monitor the battery by sending electrical pulses to thebattery and measuring its response. Then use machine learning models todiscover specific features in the electrical response that are signs of batteryaging.
The researchers conducted more than 20,000experimental measurements to train the model, which is the largest data set inits class. The researchers also show that machine learning models can beinterpreted as giving hints of degraded physical mechanisms. The model can tellwhich electrical signals are most relevant to aging, which in turn allows themto design specific experiments to explore the causes and ways of batterydegradation.
Researchers are now using their machinelearning platform to understand the degradation of different battery chemistry.They also developed the best battery charging protocol, powered by machinelearning, to achieve fast charging and minimize degradation.
It is worth noting that in addition toCambridge University, universities and enterprises including StanfordUniversity, Toyota Motor, Panasonic and others are also using AI technology todevelop new batteries and improve battery performance. For example, StanfordUniversity and Toyota researchers have developed a new machine learning methodthat is said to accelerate the development of electric vehicle batteries.Specifically, the research team of MIT Stanford and Toyota Research Institutehas developed a method based on machine learning, which shortens the batterycharging test time from nearly two years to 16 days, a reduction of nearly 15times, which helps Speed up the development of new batteries. Panasonic has also developed an AI high-tech material analysismethod that can visualize the behavior of lithium-ion battery internalmaterials during battery operation under high-speed and high-resolutionconditions. The visualization of this state will greatly affect the lithiumbattery. The capacity density, charge and discharge speed and life areimproved.
Ashlee Peng, February 10th, 2020