Online health prognosis for lithium-ion batteries under dynamic discharge conditions over wide temperature range

Abstract

Battery state of health assessment is crucial for enabling effective battery safety management and optimization control. However, battery health estimation often becomes difficult when dealing with complex operating conditions and different temperatures. In order to estimate state of health under different temperatures and dynamic operating conditions, battery experiments with multiple operating conditions are conducted. Then, three aging features—discharge voltage integration, discharge time, and net discharge energy—are extracted from the experiment data of lithium-ion batteries, along with two operating condition features—mean current and discharge capacity ratio. By fusing these features, three fused health indicators are obtained. Fused indicators are used as inputs for a Gaussian process regression model to build an accurate capacity estimation model. To account for the influence of temperature on battery health, a method for extracting health indicators over a wide temperature range is proposed. By determining the temperature baseline and the relationship equation between aging and operating condition features in the initial cycle, it is observed that the relationship between health indicators and capacity is not affected by temperature. Based on this, a battery capacity estimation model for a wide temperature range is developed. This method achieves high-precision battery health estimation and can be generalized to different operating conditions, a wide temperature range, and various battery material systems. It offers a new approach for assessing battery health under dynamic operating conditions and holds potential practical applications.