使用兩個連續且不等的方波對鋰離子電池同時進行安全和狀態評估

Abstract

鋰離子電池 (lithium ion batteries, LIBs) 因其高效能和耐用性，廣泛應用於電動車和再生能源系統。當LIBs退化且無法再提供高性能時，大規模退役這些電池會帶來顯著的環境和經濟挑戰。相比於處置和回收，再利用電池不僅延長了其壽命和價值，還減少了對新電池製造成本的需求。然而，未及早檢測到的內部短路（ISC）可能會導致熱失控和嚴重事故。現有的診斷方法為了解決上述問題，需要完整的充放電周期或長時間的靜置期以獲取足夠的訊息，檢測過程相對耗時。因此，本研究提出了一種高效的診斷方法，僅通過110秒的兩個連續方波測量即可獲得安全訊息和電池狀態。由於ISC和電池狀態對電化學阻抗譜都有影響，第一個方波主要用於評估電池狀態，而第二個方波提供識別ISC的關鍵訊息。通過利用歐姆電阻、初始電壓和低頻段的前三個阻抗作為特徵，並將其應用於機器學習模型，該方法可以準確地將ISC的嚴重程度分類到不同等級，準確率達93.83%，同時預測電池健康狀態和電量狀態，其均方根誤差分別為2.22% 和1.72%。該方法特別適用於LIBs快速的狀態評估與安全性分類，為電池再利用提供了重要價值，有助於最大化汰役電池的剩餘價值，延長其使用壽命，並促進可持續能源的利用。

LIBs are widely used in electric vehicles and renewable energy systems due to their high efficiency and durability. When LIBs degrade and can no longer deliver high performance, the large-scale retirement of these batteries poses significant environmental and economic challenges. Compared to disposal and recycling, battery reuse not only extends its lifespan and value but also reduces the demand for manufacturing costs of new batteries. However, internal short circuits (ISC) not detected early can lead to thermal runaway and severe accidents. Existing diagnostic methods to tackle the above problem, however, are time-consuming as they require complete charge and discharge cycles or long relaxation period for sufficient information to be obtained. Therefore, this study proposes an efficient diagnostic method that obtains safety information as well as battery states through only 110 seconds of two consecutive square waves measurement. Due to the influence of both ISC and battery states on the impedance spectrum, the first square wave is primarily utilized to evaluate the battery state, while the second square wave provides crucial information for identifying the ISC. By utilizing the ohmic resistance, initial voltage, and the first three low-frequency impedances as features and applying them to a machine learning model, the proposed method can accurately classify the severity of ISC into different levels with an accuracy of 93.83%, while simultaneously predicting the battery state of health and state of charge with root mean square errors of 2.22% and 1.72%, respectively. This method is particularly suitable for rapid LIBs safety and state assessment, providing significant value for battery reuse initiatives. It contributes to maximizing the residual value of decommissioned batteries, extending their useful life, and promoting sustainable energy utilization.