Accurate and Efficient SOH Estimation for Retired Batteries

There will be an increasing number of retired batteries in the foreseeable future. Retired batteries can reduce pollution and be used to construct a battery cycle ecosystem. To use retired batteries more efficiently, it is critical to be able to determine their State of Health (SOH) precisely and speedily. SOH can be estimated accurately through a comprehensive and inefficient charge-and-discharge procedure. However, the comprehensive charge and discharge is a time-consuming process and will make the SOH assessment for many retired batteries unrealistic. This paper proposes an accurate and efficient SOH Estimation (SOH-E) method using the actual data of retired batteries. A battery data acquisition system is designed to acquire retired batteries’ comprehensive discharge and charge data. The acquired discharge data are separated into various time interval-segregated sub-data. Then, the specially designed features for SOH-E are extracted from the sub-data. Neural Networks (NNs) are trained using these sub-data. The retired batteries’ SOH levels are then estimated after the NNs’ training. The experiments described herein use retired lead–acid batteries. The batteries’ rated voltage and capacity are 12 V and 90 Ah, respectively. Different feature value extractions and time intervals that might affect the SOH-E accuracy and are tested. The Backpropagation NN (BPNN) and Long-Short-Term-Memory NN (LSTMNN) are designed to estimate SOH in this paper. The experimental results indicate that SOH can be calculated in 30 min. The Root-Mean-Square Errors (RMSEs) are less than 3%. The proposed SOH-E can help decrease pollution, extend the life cycle of a retired battery, and establish a battery cycle ecosystem.