Battery

In our research, we apply various physics-based models to simulate the internal behavior of batteries. One of the core models we use is the Pseudo Two-Dimensional (P2D) model, which accurately captures complex electrochemical and transport processes, such as ion diffusion and reaction kinetics, within lithium-ion batteries. This model is essential for understanding battery performance under different conditions, but its computational complexity can be a challenge.

To balance accuracy and computational efficiency, we also use the Single Particle Model (SPM), which simplifies internal dynamics while still providing valuable insights for control and optimization. Additionally, the Extended Single Particle Model (ESPM) is employed for aging-related research, as it incorporates more detailed electrochemical mechanisms such as side reactions and degradation, making it ideal for studying long-term battery performance.

At our lab, we utilize advanced numerical techniques to reduce the complexity of these models, including the P2D model, enabling faster simulations while maintaining the precision needed for critical applications like performance optimization and degradation analysis.

These physics-based modeling approaches are not limited to conventional lithium-ion batteries (LIBs). We also apply them to cutting-edge battery technologies such as All-Solid-State Batteries (ASSBs), which promise enhanced safety and energy density. Furthermore, our models are tailored to simulate Silicon-Graphite composite anodes, addressing the challenges of high-capacity anode materials. We extend our research to Redox Flow Batteries (RFBs), optimizing their large-scale energy storage capabilities for applications such as grid storage.

In battery modeling, the process of identifying parameters is particularly challenging due to the coupling of multiple equations and the inherent nonlinearity in these systems. As a result, finding a single set of parameter values is often not feasible, and the uniqueness of solutions cannot always be guaranteed.

To address this, our approach focuses on identifying the distribution of parameters rather than pinpointing a single value. This allows us to better account for the variability and uncertainty inherent in battery behavior. We generate parameter distributions through multiple optimization runs, each yielding different but plausible parameter sets, which collectively provide a more comprehensive understanding of the system.

Additionally, we apply Bayesian inference techniques to quantify the uncertainty in our models. By incorporating prior knowledge and observed data, we determine the posterior distribution of parameters, providing a probabilistic view of their values. This approach not only enhances the robustness of our models but also allows for uncertainty quantification, which is critical for making reliable predictions about battery performance and degradation under various conditions.

By focusing on parameter distributions, we improve the reliability and predictive accuracy of our models, enabling better decision-making in battery design, optimization, and management.

Unlike physics-based models, which rely on fundamental equations to simulate battery behavior, data-driven models leverage historical and operational data to make predictions. This approach is particularly advantageous for real-time applications where quick and efficient predictions are required, such as in battery state estimation—including State of Charge (SoC) and State of Health (SoH)—and fault detection.

By training machine learning algorithms on large datasets, data-driven models can identify patterns and correlations within the data that may not be easily captured by physics-based models. This allows for fast and accurate predictions about battery performance under various conditions, often with lower computational cost. As a result, data-driven models are well-suited for applications that require real-time monitoring and decision-making, such as electric vehicles, energy storage systems, and battery management systems (BMS).

While physics-based models provide deep insights into the underlying electrochemical processes, data-driven models complement them by offering efficiency and scalability, enabling continuous and dynamic battery management in practical applications.