Without our cell phones or computers, what would our lives be like? Electronics continues to be a part of our daily routines from toys to laundry machines to electric cars.
One issue that hinders the rapid growth of battery technology is that it takes a long time to test and monitor battery 'health' that impacts battery life. Therefore, better methods are much needed to predict battery life, but they are too difficult to develop. A new approach was made to find how to accurately determine the valuable lifespan of lithium-ion batteries that are used in devices from mobile phones to electric cars in advance that could speed up the development of batteries and enhance manufacturing. Machine learning builds the models that accurately predict battery life using data collected from charging-discharge cycles computed in the early stages of the life of a battery.
To tackle the imminent climate crisis, society needs to stop producing carbon emissions. A double strategy has surfaced to achieve this objective: the electricity sector is using renewable energy sources, and electric vehicles are supplanting those using conventional combustion engines. Both changes come with barriers of their own.
One of the main obstacles for renewable energy is that the sources are often localized, creating a demand-supply imbalance. The problems with electrified transport are guaranteeing that when conventional combustion engines are no longer used, sufficient electricity is generated to charge all vehicles and the integration of charging infrastructure with the electrical grid. Therefore, the marketing of electric and hybrid vehicles has triggered an increasing demand for long-lived batteries for both driving and grid buffering. As a result, battery health assessment methods are becoming more and more crucial.
A battery's runtime prediction depends not only on the start (State of Charge) SoC but also on other factors such as battery health and imposed road profile. The self-corrective regression model is proposed and enforced in order to overcome this type of difficulties. The main issue with SoC estimation is to determine a battery's initial SoC. Comprehensive experiments are required to calculate the initial SoC and may vary with the battery life as well. Data-driven modeling using machine learning is a successful route for lithium-ion battery prognostics and therefore could help to develop, manufacture and optimize emerging battery technology.