利用非準靜態電化學阻抗譜和端電壓快速監測鋰離子電池狀態

Abstract

準確監測鋰離子電池 (lithium ion batteries, LIBs) 的健康度 (state of health, SOH) 和電池電量 (state of charge, SOC) 是至關重要，特別是在電動汽車依賴這些電池作為主要能源的情況下。電化學阻抗譜 (electrochemical impedance spectroscopy, EIS) 已成為非侵入式探索電池阻抗的強大技術，該阻抗與其兩種電池狀態直接相關。然而，由於需要全頻率測量以及電池需要長弛豫時間以達完全放鬆狀態的要求，古典 EIS 測試十分耗時。為了解決這些弱點，本研究提出了一種新方法：非準靜態 EIS 測量。該方法涉及在充電或放電過程結束時，在短暫弛豫時間後立即對 LIBs 進行 EIS 測試。通過採用這種方法，我們觀察到高頻和隨後的部分中頻阻抗對弛豫時間的影響性很低。同時，此區段的阻抗隨著電池狀態的不同而變化。利用該頻率區段內的阻抗以及端電壓，我們可以預測電池的 SOH 和 SOC。為了實現這一目標，我們採用高斯過程回歸模型 (Gaussian process regression, GPR)，以阻抗和端電壓作為輸入。值得注意的是，我們的結果表明該方法可以將輸入維度減少到 14 以下，並且獲得必要測量所需的時間可以顯著縮短到不到 7 s。此外，我們對電池 SOH 和 SOC 的預測具有很高的準確性，均方根誤差分別低於 2.66% 和 1.57%。這些發現凸顯了我們提出的方法在高效可靠地監測 LIBs 方面的潛力，從而能夠增強電動汽車應用中的性能評估和優化。通過採用非準靜態 EIS，我們可以克服與古典 EIS 測試相關的時間限制，從而有助於更快速、更準確地監測電池 SOH 和 SOC。這一進步有望推動電動汽車領域的發展並優化其能源管理系統。

Accurate monitoring of the SOH and SOC in LIBs is of utmost importance, particularly in the context of electric vehicles relying on these batteries as their primary energy source. EIS has emerged as a powerful technique for exploring the impedances of the battery, which are directly linked to its two states. However, classical EIS testing is time-consuming due to the need for broadband frequency measurements and the requirement of a full relaxation time for the battery. To address these weaknesses, this study proposes a novel approach: non-quasi-static EIS. This method involves conducting the EIS test on LIBs immediately after a short relaxation time following the charging or discharging process. By employing this approach, we have observed that the high- and subsequent partial medium-frequency impedances exhibit minimal dependence on the relaxation period. At the same time, their values vary with the battery states. Exploiting the impedance values within this frequency range, along with the terminal voltage, we can estimate the SOH and SOC of the battery. To achieve this, we employ a GPR model, with the impedance values and terminal voltage as inputs. Remarkably, our results indicate that the input dimension can be reduced to below 14, and the time required to obtain the necessary measurements can be significantly shortened to less than 7 seconds. Furthermore, our estimations of the SOH and SOC exhibit high accuracy, with root mean square errors below 2.66% and 1.57%, respectively. These findings highlight the potential of our proposed method for efficient and reliable monitoring of LIBs, enabling enhanced performance assessment and optimization in electric vehicle applications. By adopting this approach, we can overcome the time constraints associated with classical EIS testing, thereby facilitating more rapid and accurate evaluations of battery health and charge levels. This advancement holds promise for advancing the field of electric vehicles and optimizing their energy management systems.