Modeling of Battery Considering Temperature and Aging Uncertainties


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date18 Oct 2021


In recent years, with the deployment of low carbon energy systems, the energy storage system used to meet electrical energy demand plays an important role in tremendous energy-related industrial fields and applications. An accurate Battery Management System (BMS) is highly demanded to ensure the optimum use of an energy storage system, while an efficient Battery Thermal Management System (BTMS) is needed to ensure the safe use of the energy storage system. Particularly, temperature and aging are the most crucial factors that influence battery performance. This thesis provides a systematic methodology for analyzing the temperature and aging uncertainties in BMS and BTMS to enhance the performance and adaptability of battery management systems.

Firstly, an electrochemical model is built to offline estimated the long-term state of health of the Vanadium-ions battery. As the key component of Vanadium-ions battery which plays the main role in battery capacity fading, the features of different membranes are compared and summarized. Then different membranes are analyzed in the vanadium/air battery. The crossover rate of vanadium ions through the membrane and oxygen transport determines the capacity of the vanadium/air redox flow battery due to the ion diffusion and the side reactions and the electro-migration and convection. The battery's high reaction temperature also impacts the diffusion coefficient. Based on Fick's Law, mass balance for each reaction ions and reaction temperature, and Arrhenius Equation to predict the temperature effect, a dynamic modeling capacity decay can be set. Then by using the Nernst Equation, the voltage change of the battery can be calculated. Finally, the electrolyte concentration change is predicted, and battery performances with different membranes, including Nafion 115, CMS, and CMX, in different working conditions are compared using the developed dynamic model.

Subsequently, the online BMSs of a single Lithium-ions battery cell are developed. The short-term and long-term state of health (SOH) and state of charge (SOC) of (LiNixCoyMn1-x-yO2 (NCM) battery are estimated. The performances of SOC and SOH with equivalent circuit model-based methods are developed first, and an extended Kalman filter is added. Besides, the data-driven estimation model is applied, including deep learning. After that, comparison and combination of these two methods are developed, and then all these methods including extended Kalman filters, fully connected deeply network with drop methods, and the combination (extended Kalman filters - fully connected deep network with drop methods) are applied by different aged and capacity batteries. Besides the battery state of the voltage and current, the relationship between inner resistance, temperature, and capacity are also considered. Subsequently, a suggested method is promising for online state estimation of battery working at different temperatures and initial working state. Finally, an APP is developed with a clustering algorithm named decision tree to show the SOH state. Three different decision tree algorithms are compared, including Decision Tree, Random Forest Tree, and Gradient Boost Decision Tree. This APP can be installed on any computer whose system is Windows XP or higher version.

Finally, an optimized forced air-cooling BTMS for parallel-connected Lithium-ions battery packs is designed to ensure temperature uniformity. This novel design BTMS of LiFePO4 cuboid battery module is adding different additional airflow outputs in a typical U-type cooling system. The temperatures of batteries with different additional airflow outputs are simulated by a three-dimensional computational fluid dynamics (CFD) model. Because the features of these additional airflow outputs, including positions, numbers, and areas of them, influence the temperature of the battery pack simultaneously, a nine-layer fully connected deep network AI model is built to find the relationship between the temperature of the pack and these features of additional airflow outputs in BTMS. Basing on this predicted AI model, after changing the features of around seven hundred thousand BTMS, finally, a suggested one with the lowest temperature difference is suggested.