enEnglish

What Is Battery State of Charge (SOC) Estimation?

Nov 19, 2025

Leave a message

Battery State of Charge (SOC) Estimation

Common Battery Models

 

The electrochemical reaction process of power batteries is complex, influenced by numerous and uncertain factors. Mathematical modeling of this process is a multidisciplinary and multi-field problem, and has always been a key focus and challenge for both academia and industry. The input excitation (load current) and output observations (voltage and temperature) of a power battery are finite measurable parameters for the power battery management system. Accurate modeling is essential for more accurately describing the external characteristics of power batteries, designing reliable power battery state estimation algorithms, and developing optimal energy management systems for new energy vehicles. Common power battery models are mainly divided into electrochemical models, equivalent circuit models, and machine learning models.

 

(1) Electrochemical Model


In the mid-1990s, M. Doyle, T.F. Fuller, and J. Newman of the University of California, Berkeley, established a pseudo-two-dimensional (P2D) model based on the theory of porous electrodes and concentrated solutions, laying the foundation for the development of electrochemical mechanism models. This model uses a series of partial differential equations and algebraic equations to accurately describe the diffusion and migration of lithium ions inside the power battery, electrochemical reactions on the surface of active particles, Ohm's law, and charge conservation, among other electrochemical phenomena. To date, most electrochemical models are derived and developed from this model. An electrochemical model is a first-principles model that can accurately simulate not only the external characteristics of a power battery but also the distribution and changes of internal characteristics (such as the concentration of lithium ions in the electrodes and electrolyte, and the reaction overpotential, which are difficult to measure). Compared with other power battery models, electrochemical models can describe the microscopic reactions inside the power battery in greater depth and have more explicit physical meaning.

 

P2D models are versatile and scalable, applicable to battery types with different material systems, and can be developed and extended into more complex multi-field coupling models. Therefore, P2D models play an irreplaceable role in battery modeling. However, they contain complex partial differential equations and numerous electrochemical parameters, placing high demands on the computational capabilities of the Battery Management System (BMS). Currently, solving P2D models primarily employs numerical methods, such as the finite difference method, the finite element method, and the finite volume method.

 

(2) Equivalent Circuit Model

 

The equivalent circuit model uses traditional circuit elements such as resistors, capacitors, and constant voltage sources to form a circuit network to describe the external characteristics of the power battery. This model uses a voltage source to represent the thermodynamic equilibrium electromotive force of the power battery, and an RC network to describe the dynamic characteristics of the power battery. The equivalent circuit model has good applicability to various operating states of the power battery, and the state equations of the model can be derived, facilitating analysis and application. The equivalent circuit model has been widely used in new energy vehicle modeling and simulation research and model-based BMS. Figure 7-27 shows a typical equivalent circuit model of a power battery composed of n RC network structures, referred to as the n-RC model. This model consists of three parts:

 

1) Voltage Source: The open-circuit voltage of the power battery is represented by $U_{oc}$.

 

2) Ohmic Internal Resistance: The contact resistance of the power battery electrode materials, electrolyte, isolation resistor, and various components is represented by $R_o$.

 

3) RC Network: The dynamic characteristics of the power battery, including polarization characteristics and diffusion effects, are described by the polarization resistance $R_p$ and polarization capacitance $C_p$, where $i = 0, ..., n_s$.

 

In Figure 7-27, Up represents the polarization voltage of the power battery.

 

Figure 7-27  n-stage RC model circuit structure

 

Based on Kirchhoff's voltage law and current law, and the relationship between capacitor voltage change and current, the state-space equation of the circuit model can be expressed as:

 

7-50

 

Commonly used power battery equivalent circuit models such as the Rint model, Thevenin model, and Dual Polarization (DP) model are special cases of the n-RC equivalent circuit model when n=0, n=1, and n=2, respectively, and have been widely used in power battery state estimation and management algorithms.

 

(3) Machine Learning Models

 

Machine learning models do not require knowledge of the battery's internal composition and specific reaction mechanisms; they only need to obtain the battery's historical operating data (current, voltage, temperature, etc.). Essentially, they establish nonlinear mapping functions between variables through data-driven methods. The main advantage of this type of model is its applicability to different battery types, its good versatility, and its ability to fully simulate the nonlinear characteristics of battery behavior.

 

In the field of power battery management and control, the machine learning methods used mainly include fuzzy logic, neural networks, support vector machines, and their combined algorithms. In March 2016, AlphaGo's victory over Go world champion Lee Sedol injected new vitality into deep learning, triggering a new wave of research and application, which has also been applied to battery management. With sufficient battery data for training, this type of model can achieve good predictive performance. However, this model lacks physical meaning, is uninterpretable, and its performance is greatly affected by the quantity and quality of training data, making it difficult to guarantee its reliability and robustness when applied to battery management systems.

Send Inquiry