Researchers Jiang Liu, Yan Qin, Wei Dai, and Chau Yuen from the School of Electrical and Electronic Engineering at Nanyang Technological University in Singapore have developed a new method for monitoring the state-of-health (SOH) of lithium-ion batteries, particularly for use in unmanned air vehicles (UAVs). Their work, published in the journal IEEE Transactions on Industrial Electronics, focuses on improving the efficiency and accuracy of SOH monitoring, which is crucial for the safe and effective operation of battery-powered devices.
Lithium-ion batteries are widely used in various portable and mobile devices, including UAVs. Accurate SOH monitoring helps in predicting the remaining useful life of the battery, ensuring optimal performance and safety. However, traditional methods of SOH monitoring require substantial computational resources, which can reduce the working endurance of portable devices.
To address this issue, the researchers proposed a lightweight transfer learning (TL)-based approach called constructive incremental transfer learning (CITL). Transfer learning is a technique that leverages knowledge from data-rich source conditions to improve learning in target conditions with limited data. The CITL method uses unlabeled data from the target domain to minimize monitoring residuals through an iterative process of adding network nodes.
The researchers also ensured the cross-domain learning ability of node parameters in CITL through structural risk minimization, transfer mismatching minimization, and manifold consistency maximization. They provided a convergence analysis of the CITL to theoretically guarantee its effectiveness and network compactness.
The proposed approach was tested using a realistic UAV battery dataset collected from dozens of flight missions. The results showed that CITL outperformed several other methods, including SS-TCA, MMD-LSTM-DA, DDAN, BO-CNN-TL, and AS^3LSTM, in terms of SOH estimation accuracy, as evaluated using the root mean square error index.
This research has significant implications for the energy sector, particularly in the development of more efficient and accurate battery monitoring systems for portable and mobile devices. By reducing the computational resources required for SOH monitoring, the proposed method can help extend the working endurance of devices like UAVs, making them more reliable and effective in various applications. The research was published in the IEEE Transactions on Industrial Electronics, a reputable journal in the field of industrial electronics and applications.
This article is based on research available at arXiv.