University of Kansas develops machine learning technology for thermal runaway of lithium-ion batteries

According to foreign media reports, mobile phones, electronic devices, laptop computers, electric vehicles such as Toyota and Tesla, Boeing 787 aircraft, and ships of the U.S. Navy are now powered by lithium-ion batteries. Although such batteries are widely used, they also have many risks. Because of the high reactivity of lithium metal, such batteries and battery cells will experience 'thermal runawayAlthough accidents are rare, they still arouse public concern about lithium-ion technology.

Now, with the support of a new 5-year, $500,000 grant from the National Science Foundation, the University of Kansas ) Researchers have developed a technology that can monitor and prevent lithium-ion batteries from overheating. Huazhen Fang, an assistant professor in the mechanical engineering department of the university, and his students developed a machine learning method to monitor the temperature inside the battery.

According to researchers at the University of Kansas, most of the current technologies for tracking the temperature of lithium-ion batteries are not mature enough, because the sensor can only read the surface temperature of the battery.

Professor Fang said: “Usually, the temperature on the surface of the battery is not enough to tell us about the state of the battery, and the temperature inside the battery tells us more about thermodynamics. But Now, there are very few ways to place sensors in the battery. However, using artificial intelligence and machine learning, we can predict the temperature inside the battery cell, allowing us to detect battery behavior. The temperature of the battery surface can provide a machine learning method A wealth of data, combined with mathematical models, can predict what is happening inside the battery.'

The battery temperature is not assumed to be a uniform temperature. Nowadays, there is a modeling method called 'lumped parameter modelProfessor Fang said that his computer learning technology can predict the temperature change inside the battery, which is a more accurate and realistic method to calculate the possibility of thermal runaway of the battery.

Professor Fang said: “When the battery is charged and discharged, the temperature distribution is uneven. Usually, the internal temperature of the battery near the electrode is higher, but the outer surface temperature is lower. The lumped parameter model only takes into account the uniform distribution of battery temperature, while our method reconstructs the battery temperature in time and space.'

Researchers at the University of Kansas The data of the ion battery is input into artificial intelligence to infer the internal temperature of the battery. Such data can be processed in battery-powered devices or connected to cloud computing. If the battery is thermally out of control, the device can be programmed to disconnect the battery when it is turned off to prevent the battery from becoming hot, catching fire or causing an explosion.

With the above innovations, lithium-ion batteries can be extended to more industrial applications by tying hundreds of batteries together. According to Professor Fang, lithium-ion technology is increasingly being used in large-scale power grids to store and discharge electricity generated by sustainable technologies such as solar and wind energy.