FormCap Model Enables Intelligent Capacity Prediction and Grading of Lithium-Ion Batteries during the Formation Process

As a key quality control step in the manufacturing of lithium-ion batteries (LIBs), capacity grading aims to screen out low-performance batteries to ensure battery pack consistency. To break through the reliance of conventional capacity grading on high-energy and high-cost cycling tests, a formation capacity intelligent capacity grading technology (FormCap-ICGT) is proposed, which realizes capacity prediction and grading by using only the formation data. Firstly, based on multistep cleaning and multilevel feature extraction of the formation data, a FormCap model that integrates serialized feature selection and attention mechanisms is constructed. It dynamically selects key capacity-related features to enable highly accurate capacity prediction. Based on this, the task of capacity grading is done in the formation process. Experiments involving tens of thousands of cells from different batches show that the average absolute error in capacity prediction is 0.19 Ah, and the accuracy of capacity grading reaches 91%. FormCap-ICGT provides a technological route to simplify or even remove the conventional capacity grading test. It is of great practical significance for reducing costs and improving production efficiency in the battery manufacturing industry.