Machine learning models can more accurately estimate the charge level of lithium-ion batteries
Two professors at Columbia University (Columbia), Matthias Preindl and AlanWest, are developing a machine learning model that can more accurately estimate the charge level of lithium-ion batteries.
According to foreign media reports, electric vehicles are driven by rechargeable lithium-ion batteries (LIB), but at present, people have not fully understood and perfected lithium-ion batteries. Given that electric vehicles are expected to replace fuel vehicles, any research that can improve the performance of lithium-ion batteries will benefit the development of electric vehicles and improve the environment. Two professors at Columbia University, Matthias Preindl and AlanWest, are developing a machine learning model that can more accurately estimate the charge level of lithium-ion batteries. At present, it is estimated that the battery charge state still has an error rate of 5%. The model developed by the team aims to reduce the error rate to 1%. The research was funded by the Columbia Data Science Institute (DataScience Institute) seed fund.
As everyone knows, the battery management system is mainly used to capture the health of the battery and predict the remaining life in the period. The above two concepts can help electric vehicle owners know when they should stop to charge their batteries and when they should arrange for battery replacement. Then, a model with high estimation accuracy allows the battery management system to identify and protect weak batteries, thereby extending the life of the battery pack.
To design the machine learning model, the team will apply disturbance signals (a series of current signals generated by power electronic converters) to lithium-ion batteries. This series of signals Allows the battery to emit a detectable electrical response. The team will test such batteries in a laboratory and use power electronic converters to obtain data from batteries installed in electric vehicles. This type of data is generated once every minute and can measure battery functions such as battery temperature, voltage, and current fluctuations, resulting in hundreds of thousands of data points. Therefore, the team is designing an algorithm to evaluate such data and design an optimization model.
Preindl is an expert on how batteries interact with external components, and chemical engineer AllenWest understands the chemical composition inside batteries. The two of them are combining their engineering knowledge and advanced data science technology to design a model that can predict how to get the best performance from current lithium-ion batteries.
Preindl said: 'In fact, we have no quantitative method to understand the behavior of lithium-ion batteries. Once we have quantitative data, we will know when the battery needs Charging, how long can it be used, when it needs to be replaced, and how to extend the life of the battery.'