A quick and intelligent screening method for large-scale retired batteries based on cloud-edge collaborative architecture

Abstract

Rapid and accurate sorting consistent cells is indispensable for the cascade utilization of retired batteries. However, the conventional full charge-discharge capacity testing methods are time-consuming and low efficient, which cannot be employed to screen large-scale retired batteries. Moreover, most existing artificial intelligence techniques sort retired batteries at cloud servers, which leads to tremendous burdens of transmission. Therefore, a fast and intelligent screening approach based on a Light Gradient Boosting Machine is proposed, which only utilizes partial charge/discharge curves, significantly reducing the test time. Meanwhile, a cloud-edge collaborative framework for separating large-scale retired batteries is developed. Specifically, the model is trained in the cloud and screens retired batteries in the edge plane. The experimental results show that the data transmission time of the cloud-edge collaborative framework is reduced by about 3 orders of magnitude than the cloud computing plane. Besides, the method achieves a 97 % classification accuracy rate. More importantly, the screening efficiency of the proposed approach is approximately 6 times higher than that of the conventional full charge-discharge testing methods. Additionally, the voltage consistency of the new module regrouped is significantly improved compared with the original module. The mean of the standard deviation is reduced by approximately 3 times.