Multifeatured entropy model for safety early warning of lithium battery

The article proposes a safety evaluation strategy based on multi-feature fusion entropy to address the limited generalizability of existing battery safety warning methods. Through adaptive dynamic weighting, the contribution of various features to battery safety is updated, while a three-level safety boundary framework is applied to assess battery safety. For a normal battery, its entropy value remains within a specified range and does not exceed the threshold. When battery safety entropy crosses the first boundary, long-term monitoring is recommended. If it crosses the second boundary, a safety alert is issued, and the alert level is raised depending on the degree of boundary exceedance. This method effectively reduces the issue of abnormal features being masked by others in multi-feature coupling scenarios.