The widespread adoption of mobile energy storage systems in applications ranging from electric grid support to portable power solutions has placed significant emphasis on the reliability and safety of their core component: the energy storage battery. Ensuring the safe, stable, and long-term operation of these battery systems is a paramount challenge. Traditional monitoring methods, often reliant on complex physical models, can struggle with accuracy under real-world, dynamic operating conditions. To address this, Intelligent Battery Management System (BMS) technology has emerged as a critical solution. This article, based on intelligent BMS technology, proposes a comprehensive safety estimation model for mobile energy storage batteries. We analyze key technical parameters, safety characteristics, and the application framework of intelligent BMS. Furthermore, we establish a battery model and a dual extended Kalman filter-based state monitoring model to simultaneously estimate the State of Charge (SoC), internal resistance, capacity, and State of Health (SoH) of the energy storage battery. Simulation results demonstrate the enhanced accuracy and effectiveness of the proposed model for battery safety state monitoring.
1. Technical Parameters and Safety Characteristics of Mobile Energy Storage Batteries
The performance and safety of a mobile energy storage system are intrinsically linked to the characteristics of its energy storage battery. Among various chemistries, Lithium Iron Phosphate (LFP) batteries are widely adopted for mobile applications due to their favorable balance of capacity, cycle life, and thermal stability. The key parameters defining an energy storage battery are summarized below.
| Parameter | Symbol/Unit | Description and Safety Implication |
|---|---|---|
| Capacity | Q (Ah) or E (kWh) | Total electrical charge or energy the energy storage battery can store. Directly relates to system runtime. Over-discharge below safe limit damages cells. |
| State of Charge | SoC (%) | Ratio of remaining capacity to maximum capacity. Accurate SoC estimation is crucial to prevent over-charge and over-discharge of the energy storage battery. |
| Open Circuit Voltage | OCV (V) | Voltage of the energy storage battery at rest. It has a non-linear relationship with SoC and is vital for model-based estimation. |
| Internal Resistance | Rint (Ω) | Opposition to current flow within the energy storage battery. Increases with aging and affects efficiency, heat generation, and terminal voltage. |
| Charge/Discharge Rate | C-rate (h⁻¹) | Rate at which the energy storage battery is charged or discharged relative to its capacity. High rates can lead to excessive heat, lithium plating, and accelerated aging. |
| Operating Temperature | T (°C) | Ambient and core temperature of the energy storage battery. Extreme temperatures can trigger thermal runaway, a critical safety failure. |
Safety incidents in energy storage batteries, such as thermal runaway, fire, or explosion, are typically triggered by a combination of electrical, thermal, and mechanical abuse. The primary safety-influencing factors and their management through system design are analyzed below.
- Electrical Abuse (Overcharge, Over-discharge, Short Circuit): Exceeding voltage or current limits causes internal heating, gas generation, and electrode degradation. A robust BMS must implement strict voltage and current protection.
- Thermal Abuse: High external temperature or internal heat generation from high currents or failure can initiate exothermic reactions. Thermal management systems (cooling/heating) and BMS temperature monitoring are essential for the energy storage battery.
- Mechanical Abuse (Crush, Penetration): Can cause internal short circuits. This is mitigated by robust module and pack design with proper spacing and protection for the energy storage battery cells.
- Internal Aging and Faults: Gradual increase in internal resistance and capacity fade can shift operating points and increase heat generation. Detection of these changes is a key function of an intelligent BMS.
The safety of a lithium-based energy storage battery system is therefore a systemic property, dependent on cell chemistry (e.g., LFP’s stable olivine structure), module design (for thermal isolation and mechanical integrity), and crucially, the performance of the Battery Management System (BMS).
2. Intelligent Battery Management System (BMS) Technology
The Intelligent Battery Management System is the central brain responsible for ensuring the safe, efficient, and reliable operation of the energy storage battery pack. It transcends basic protection by incorporating advanced algorithms, models, and communication for state estimation and predictive management.
2.1 Core Technical Principles
The intelligent BMS operates on a closed-loop principle of measurement, estimation, and control. Its technical foundation includes:
- High-Precision Parameter Acquisition: Simultaneously samples voltage of individual cells or modules, total pack current, and temperature at multiple critical points within the energy storage battery pack.
- Advanced State Estimation: Employs algorithmic models (e.g., Kalman Filters, neural networks) to estimate non-measurable states like SoC and SoH in real-time, using the acquired voltage, current, and temperature data.
- Model-Based Prediction and Control: Uses the estimated states and a battery model to predict future behavior, enabling proactive control strategies such as dynamic current limiting, adaptive charging, and thermal management for the energy storage battery.
- Integrated Communication: Facilitates data exchange with external controllers (e.g., energy management system) and user interfaces, allowing for remote monitoring, configuration, and diagnostics of the energy storage battery system.
2.2 Functional Architecture and Role in Safety
A typical intelligent BMS for a mobile energy storage battery comprises several key functional modules, each contributing to system safety.
| BMS Functional Module | Primary Safety Functions |
|---|---|
| Measurement & Acquisition | Provides the raw data (V, I, T) essential for all safety algorithms and protection triggers. |
| State Estimation (SoC/SoH) | Prevents hidden over-charge/discharge by providing accurate capacity awareness. SoH tracks aging, signaling when the energy storage battery needs maintenance or replacement. |
| Cell Balancing | Equalizes charge among cells, preventing any single cell from being overstressed, which extends pack life and improves safety. |
| Protection Logic | Executes immediate hardware-based disconnection in case of detected over-voltage, under-voltage, over-current, or extreme temperature. |
| Thermal Management Control | Activates cooling or heating systems based on temperature readings to keep the energy storage battery within its safe operating window. |
| Fault Diagnosis & Logging | Identifies and records anomalies (e.g., sudden internal resistance change, isolation fault), enabling preventive maintenance. |
The intelligent BMS acts as the primary guardian against the safety factors listed in Section 1. For instance, its precise SoC estimation directly mitigates overcharge/over-discharge risks. By continuously monitoring temperature and controlling thermal systems, it manages thermal abuse. Furthermore, by tracking the incremental increase in the internal resistance of the energy storage battery, it can predict potential failure points before they become critical.
3. Battery Modeling and Safety State Estimation Framework
To enable model-based state estimation within the intelligent BMS, an accurate yet computationally efficient representation of the energy storage battery’s dynamics is required.
3.1 Equivalent Circuit Model (ECM)
A second-order Thevenin model, also known as a Dual Polarization (DP) model, is widely used for its good balance between accuracy and complexity. This model represents the energy storage battery with the following components:
- An Open Circuit Voltage (OCV) source, which is a function of SoC: $V_{ocv}(z)$.
- A series resistance $R_0$, representing instantaneous voltage drop.
- Two parallel $R-C$ pairs ($R_1$,$C_1$ and $R_2$,$C_2$) modeling the short-term and long-term transient voltage responses, respectively.
The state-space equations for this energy storage battery model are discretized as follows:
State Equation:
$$
\begin{bmatrix}
V_1(k+1) \\
V_2(k+1) \\
z(k+1)
\end{bmatrix} =
\begin{bmatrix}
\alpha_1 & 0 & 0 \\
0 & \alpha_2 & 0 \\
0 & 0 & 1
\end{bmatrix}
\begin{bmatrix}
V_1(k) \\
V_2(k) \\
z(k)
\end{bmatrix}
+
\begin{bmatrix}
\beta_1 \\
\beta_2 \\
– \frac{\eta \Delta T}{Q_n}
\end{bmatrix}
I(k) + w(k)
$$
where $\alpha_i = \exp(-\Delta T / (R_i C_i))$, $\beta_i = R_i(1-\exp(-\Delta T / (R_i C_i)))$, $z$ is SoC, $I$ is current (positive for discharge), $\eta$ is coulombic efficiency, $\Delta T$ is sampling time, $Q_n$ is nominal capacity, and $w(k)$ is process noise.
Output Equation:
$$
V_t(k) = V_{ocv}(z(k)) – I(k)R_0 – V_1(k) – V_2(k) + v(k)
$$
where $V_t$ is the terminal voltage of the energy storage battery and $v(k)$ is measurement noise.
The OCV-SoC relationship $V_{ocv}(z)$ is often characterized by a polynomial or lookup table. The model parameters $\theta = [R_0, R_1, C_1, R_2, C_2, Q_n]$ can be identified offline from test data using optimization algorithms that minimize the error between model voltage and measured voltage:
$$
\min_{\theta} \sum_{k=1}^{N} (V_t^{meas}(k) – V_t^{model}(k, \theta))^2
$$
3.2 Dual Extended Kalman Filter (DEKF) for Joint State-Parameter Estimation
A key challenge for safety monitoring is that the model parameters, especially the internal resistance $R_0$ and actual capacity $Q$, change as the energy storage battery ages. A single filter estimating only SoC with fixed parameters will lose accuracy. The Dual Extended Kalman Filter (DEKF) provides an elegant solution by running two intertwined EKFs in parallel: one for state estimation (EKF-x) and one for parameter estimation (EKF-θ).
EKF-x: State Estimation Filter
This filter estimates the dynamic states $x_k = [V_1(k), V_2(k), z(k)]^T$ assuming known parameters $\theta_k$.
$$ \hat{x}^-_k = f(\hat{x}_{k-1}, I_{k-1}, \hat{\theta}_{k-1}) $$
$$ P^-_x = A_k P_{x,k-1} A_k^T + Q_x $$
$$ K_x = P^-_x H_k^T (H_k P^-_x H_k^T + R_x)^{-1} $$
$$ \hat{x}_k = \hat{x}^-_k + K_x (V_t^{meas}(k) – \hat{V}_t(k)) $$
$$ P_{x,k} = (I – K_x H_k) P^-_x $$
where $A_k$ and $H_k$ are Jacobian matrices of the state and output functions, $Q_x$ and $R_x$ are noise covariance matrices for the state process and measurement, and $K_x$ is the Kalman gain for states.
EKF-θ: Parameter Estimation Filter
This filter treats the key variable parameters as a state vector, e.g., $\theta_k = [R_0(k), Q(k)]^T$. They are modeled as a random walk process.
$$ \hat{\theta}^-_k = \hat{\theta}_{k-1} $$
$$ P^-_\theta = P_{\theta,k-1} + Q_\theta $$
The parameter update uses the state estimation error. The sensitivity of the terminal voltage prediction error to parameter changes is calculated.
$$ K_\theta = P^-_\theta J_k^T (J_k P^-_\theta J_k^T + R_\theta)^{-1} $$
$$ \hat{\theta}_k = \hat{\theta}^-_k + K_\theta (V_t^{meas}(k) – \hat{V}_t(k)) $$
$$ P_{\theta,k} = (I – K_\theta J_k) P^-_\theta $$
where $J_k$ is the Jacobian relating the terminal voltage error to parameter changes, and $Q_\theta$, $R_\theta$ are the corresponding noise matrices.
The two filters run concurrently. At each time step, EKF-x uses the latest parameter estimates $\hat{\theta}_{k-1}$ to predict states, and EKF-θ uses the latest state estimates $\hat{x}_k$ to update the parameters. This allows for real-time tracking of both the SoC of the energy storage battery and its evolving internal resistance and capacity.
3.3 Derivation of Safety Indicators
From the DEKF outputs, primary safety indicators for the energy storage battery are derived:
- State of Charge (SoC): Directly provided by the state vector $\hat{z}_k$ from EKF-x.
- State of Health (SoH): Can be defined in terms of capacity fade or power fade (resistance increase).
- Capacity-based SoH: $SoH_Q(k) = \frac{\hat{Q}(k)}{Q_{nominal}} \times 100\%$ where $\hat{Q}(k)$ is the estimated capacity from EKF-θ.
- Resistance-based SoH: $SoH_R(k) = \frac{R_{EoL} – \hat{R}_0(k)}{R_{EoL} – R_{0,new}} \times 100\%$. Here, $R_{0,new}$ is the initial resistance, and $R_{EoL}$ (e.g., $1.6 \times R_{0,new}$) is the resistance at end-of-life.
A comprehensive safety assessment within the intelligent BMS would monitor these indicators against thresholds, trigger alarms for low SoC or rapidly degrading SoH, and can even inform derating strategies to prolong the safe life of the energy storage battery.
4. Simulation Analysis and Discussion
To validate the proposed DEKF-based safety monitoring framework, a simulation study was conducted using a model of a 50 kWh LFP energy storage battery, typical for mobile grid-support applications. A standard urban driving/duty cycle profile was used to generate dynamic current loads.
4.1 SoC Estimation Accuracy
The performance of the proposed DEKF was compared against a standard EKF with fixed parameters. The table below shows the Root Mean Square Error (RMSE) of SoC estimation over multiple charge-discharge cycles.
| Cycle Number | SoC Estimation RMSE (%) | |
|---|---|---|
| Standard EKF (Fixed Params) | Proposed DEKF (Adaptive) | |
| 1 | 1.42 | 0.89 |
| 2 | 1.87 | 1.05 |
| 3 | 2.35 | 1.21 |
The DEKF consistently shows lower error. This is because the standard EKF’s accuracy degrades as the actual internal resistance of the energy storage battery increases with simulated aging, while the DEKF adapts by tracking $R_0$ and $Q$ in real-time.
4.2 Tracking of Aging Parameters and SoH
The DEKF’s ability to track time-varying parameters is crucial for health monitoring. The following table shows the filter’s estimation of capacity and internal resistance against their pre-defined degradation trends over cycles, and the derived $SoH_Q$ and $SoH_R$.
| Simulated Cycle Block | True Capacity (Ah) | Estimated Capacity (Ah) | $SoH_Q$ (%) | True $R_0$ (mΩ) | Estimated $R_0$ (mΩ) | $SoH_R$ (%) |
|---|---|---|---|---|---|---|
| Initial (0) | 140.0 | 139.8 | 99.9 | 30.0 | 30.1 | 99.8 |
| Mid-Life (150) | 133.0 | 132.5 | 94.6 | 36.0 | 36.3 | 89.5 |
| EoL Threshold (300) | 126.0 | 125.2 | 89.4 | 45.0 | 44.7 | 74.8 |
The results demonstrate that the DEKF-based framework effectively tracks the degradation of the energy storage battery, providing accurate estimates of both capacity and resistance-based health indicators. The $SoH_R$ value drops more sharply, highlighting resistance increase as a leading indicator of power fade and potential thermal issues. An intelligent BMS can use this information to warn operators or limit the maximum discharge current of the energy storage battery to maintain safety as it ages.
5. Conclusion
This research presents a comprehensive safety monitoring framework for mobile energy storage batteries, centered on the application of intelligent BMS technology. By moving beyond simple voltage and temperature protection, the proposed approach utilizes a high-fidelity battery model and a Dual Extended Kalman Filter algorithm to perform real-time, joint estimation of the critical states and parameters of the energy storage battery—namely, the State of Charge, internal resistance, and capacity.
The simulation analysis confirms the superiority of this adaptive method. Compared to a standard filter with fixed parameters, the DEKF maintains higher accuracy in SoC estimation as the battery ages. More importantly, it successfully tracks the gradual increase in internal resistance and decrease in capacity, enabling the direct calculation of State of Health metrics. This capability allows the intelligent BMS to transition from reactive protection to predictive health management. By identifying trends in resistance and capacity fade, the system can forecast potential safety risks associated with an aging energy storage battery, such as increased heat generation during high loads, and take preemptive action. This significantly enhances the operational safety, reliability, and longevity of mobile energy storage systems, ensuring that the energy storage battery remains a safe and dependable asset throughout its service life.