Real-Time Temperature Field Monitoring for Battery Thermal Processes: A Dual-Decomposition-Based Approach

Abstract

Real-time temperature field estimation plays a vital role in ensuring the safe operation of lithium-ion batteries and has therefore attracted considerable interest from both industrial and academic communities. However, due to the restrictions on the number of sensors in practical applications, accurate and effective estimation of the temperature field remains challenging. In this paper, a dual-decomposition-based approach is proposed to address this issue using only a limited number of temperature sensors. The key novelty lies in the simultaneous decomposition of both the state variable and the heat source term into time-space decoupled forms, which enables the design of a computationally efficient lower-order adaptive observer. An adaptive learning algorithm is further developed to estimate the unknown source coefficients online without requiring full-state measurements. Theoretical analysis confirms that the estimation errors are uniformly ultimately bounded. Extensive experiments on a real cylindrical lithium-ion battery validate the effectiveness of the proposed method, demonstrating its potential for real-time battery thermal monitoring in practical applications. Note to Practitioners—This manuscript presents a simple yet effective method for real-time temperature field monitoring in battery thermal processes using a dual-decomposition-based approach. A major advantage of the proposed method is that it requires significantly fewer sensors than existing approaches, making it particularly suitable for practical battery management systems where sensor installation is costly and space is limited. In addition, the method is straightforward to implement and involves only a limited number of tuning parameters, which facilitates its deployment in real-world applications. The adaptive learning mechanism enables online estimation of both the temperature distribution and the internal heat generation rate, providing valuable information for thermal fault detection and safe operation. These features make the proposed method especially attractive for electric vehicle battery packs and large-scale energy storage systems where real-time thermal monitoring is essential.