As the core energy supply unit of seismic stations, the energy storage cell pack plays a critical role in ensuring green, economical, safe, and stable operation. This review addresses the state-of-the-art online monitoring technologies for energy storage cell status in seismic stations, leveraging intelligent algorithms. We introduce several existing online monitoring systems, relatively mature modeling and analysis methods for energy storage cells, and the current application status of battery management systems (BMS) in seismic stations. By analyzing parameters of different energy storage cells and discussing their corresponding diagnostic difficulties, we systematically compare existing online monitoring devices and their underlying methods, drawing from both domestic and international literature.
Keywords: energy storage cell, SOC estimation, battery monitoring system, balanced control, seismic station
1. Introduction
The National Earthquake Intensity Rapid Reporting and Early Warning Project in China, scheduled for completion by the end of 2023, aims to achieve second-level earthquake early warning and minute-level intensity reporting in key regions. The project involves building 15,391 stations, including 1,928 benchmark stations and 3,114 basic stations, each equipped with 6 batteries, totaling over 30,000 energy storage cells. Continuous observation data are essential for seismic monitoring, rapid reporting, and scientific research. The power supply system, often AC-based with UPS backup, must rely on energy storage cells to ensure uninterrupted operation during grid failures. Even in remote stations using solar or wind power, energy storage cells remain indispensable. Thus, the performance of the energy storage cell pack directly impacts the reliability of the entire seismic early warning system.
Given the critical nature of these energy storage cells, real-time monitoring of their status—such as state of charge (SOC), voltage, temperature, and internal resistance—is paramount. This review provides a comprehensive overview of online monitoring technologies for energy storage cell status, focusing on lithium-ion batteries, SOC estimation algorithms, balancing control methods, and their application in seismic stations.
2. Development of Lithium-ion Batteries and SOC Estimation Research
2.1 Overview of Lithium-ion Batteries
Lithium-ion batteries dominate modern energy storage due to their high energy density (100–270 mAh/g, 5–7 times that of lead-acid batteries), high nominal voltage (3.0–4.2 V per cell), low self-discharge rate (below 1% per month at normal temperature), long cycle life (over 6 years, with LiFePO4 cells exceeding 10,000 cycles), compact size, light weight, no memory effect, wide operating temperature range, and environmental friendliness. However, they also present drawbacks: high production cost, potential safety hazards under extreme conditions, and damage from overcharge or overdischarge.
The history of lithium-ion batteries dates back to the 1970s when M.S. Whittingham created the first lithium battery using titanium sulfide as the cathode. In the early 1980s, researchers at the Illinois Institute of Technology discovered that embedding lithium into graphite improved safety. Bell Labs later developed a cathode based on intercalated lithium. In the mid-1990s, Padhi introduced LiFePO4 as a cathode material, significantly enhancing performance. In 1992, Sony commercialized the first lithium-ion battery, revolutionizing portable electronics. Today, lithium-ion cells are connected in series or parallel to meet voltage and current requirements, as shown conceptually in the connection schematic (typical series/parallel arrangement).
2.2 SOC Estimation and Balancing Control
Accurate SOC estimation is fundamental for energy storage cell management. Numerous studies have explored various methods. For instance, some researchers improved the PNGV equivalent circuit model by considering multiple battery parameters and combined it with the extended Kalman filter (EKF) to correct SOC estimates, achieving higher accuracy. Others employed iterative EKF to enhance estimation precision. A common approach involves using the Thevenin equivalent circuit model, identifying parameters offline, and applying EKF with adaptive measurement noise covariance to improve robustness under varying currents.
Random forest (RF) algorithms have been used to evaluate the importance of input features, followed by extreme learning machines (ELM) optimized by sparrow search algorithm (SSA) for SOC prediction. This hybrid method demonstrated superior accuracy under dynamic driving cycles. Additionally, dual Kalman filters have been proposed to optimize SOC for wind power smoothing, where one filter handles grid power targets and the other manages energy storage cell SOC to maintain operational limits.
Balancing control is equally critical. Traditional passive balancing dissipates excess energy, while active balancing transfers energy between cells. A two-layer distributed balancing management system uses voltage differences to calculate balancing steps, improving efficiency. Novel switched-capacitor circuits overcome performance degradation in long series strings, requiring only complementary square-wave signals. Distributed control strategies based on sparse communication networks enable SOC balancing among modules in reconfigurable energy storage systems. Hybrid active balancing methods, combining transformer-based circuits with inductor-based circuits controlled by dedicated chips, have shown faster balancing speed and higher efficiency compared to pure inductance or pure active methods.
Table 1 summarizes key SOC estimation techniques and their characteristics.
| Method | Model/Algorithm | Advantages | Limitations |
|---|---|---|---|
| Improved PNGV + EKF | Equivalent circuit model, Extended Kalman filter | High accuracy under dynamic conditions | Computational complexity |
| Iterative EKF | Thevenin model, iterative Kalman filter | Reduced estimation error | Sensitive to initial conditions |
| RF + SSA-ELM | Random forest feature selection, Sparrow search optimized ELM | Fast training, robust prediction | Requires extensive training data |
| Dual Kalman filter | Two cascaded Kalman filters | Real-time SOC and grid power coordination | Higher computational demand |
| AUKF-BP neural network | Adaptive unscented Kalman filter + BP neural network | Good noise robustness | Network training overhead |
| BAS-BP neural network | Beetle antennae search optimized BP | Improved convergence speed | May not handle all dynamic profiles |
Table 2 compares various active balancing control methods.
| Method | Topology | Balancing Speed | Efficiency | Complexity |
|---|---|---|---|---|
| Two-layer distributed | Voltage-difference based step adjustment | Moderate | High | Moderate |
| Zero-current switching switched-capacitor | Switched-capacitor network | Fast | High | High |
| Distributed control for reconfigurable systems | Communication-based SOC sharing | Moderate | High | Moderate |
| Hybrid active (transformer + inductor) | Transformer + dedicated chip (BQ78PL116) | Fast | Very high | High |
3. Research Status of Energy Storage Cell Monitoring Systems
3.1 Domestic and International Research Progress
Globally, battery monitoring systems have evolved since the 1980s. In 1989, the Electric Power Research Institute in the US, in collaboration with National Power Research Corporation, developed the PBWC lead-acid battery online monitoring system, which used multiple sensors to measure overall and individual cell voltages, current, internal temperature, and ambient temperature, transmitting data via optical cables. In the late 1990s, low-voltage stress AC-linked charge equalizing systems for VRLA batteries were proposed.
The 21st century saw a surge in BMS development. The US company Aerovironment introduced the Smart Guard system, a highly integrated BMS that allowed some battery cells to discharge while others charged via photovoltaic sources. Germany’s Hauck designed the BATTMAN system, which could manage various battery types through hardware-software combinations, improving BMS versatility. Preh GmbH developed BMS with independent protection chips for extreme operating conditions. In 2018, researchers at the Technical University of Munich proposed a cloud-connected BMS that continuously updated battery state on the server side, calculating residual values using empirical aging models. In 2020, Texas A&M University developed a cloud-based platform for large-scale lithium-ion BMS, using IoT devices and Google Cloud for data storage and analysis. Aachen University’s ISEA institute introduced a cloud BMS leveraging digital twin technology for SOC and SOH estimation.
In China, many institutions have contributed. Some researchers used relays to switch between cells for voltage measurement, but this approach suffered from slow response and limited relay lifespan. Another design implemented parallel battery unit current measurement for fault diagnosis, though it was not applicable to series-connected cells. A high-voltage analog switch combined with a differential operational amplifier (MAX14752 + INA148UA) provided accurate voltage detection for 15-cell series strings, enabling precise balancing and SOC estimation.
Mathematical models commonly used in energy storage cell monitoring include the equivalent circuit model (ECM) and electrochemical models. For ECM, the Thevenin model is popular. The state-space representation is:
$$ \begin{bmatrix} \dot{V}_1 \\ \dot{V}_2 \end{bmatrix} = \begin{bmatrix} -\frac{1}{R_1 C_1} & 0 \\ 0 & -\frac{1}{R_2 C_2} \end{bmatrix} \begin{bmatrix} V_1 \\ V_2 \end{bmatrix} + \begin{bmatrix} \frac{1}{C_1} \\ \frac{1}{C_2} \end{bmatrix} I_L $$
$$ V_t = V_{oc}(SOC) – V_1 – V_2 – I_L R_0 $$
where $V_1$, $V_2$ are polarization voltages, $R_1$, $C_1$, $R_2$, $C_2$ are polarization parameters, $R_0$ is the ohmic resistance, $V_{oc}$ is open-circuit voltage as a function of SOC, and $I_L$ is the load current.
For SOC estimation using the extended Kalman filter, the discrete-time equations are:
$$ x_{k+1} = f(x_k, u_k) + w_k $$
$$ y_k = h(x_k, u_k) + v_k $$
where $x_k = [SOC_k, V_{1,k}, V_{2,k}]^T$, $u_k = I_{L,k}$, $y_k = V_{t,k}$, and $w_k$, $v_k$ are process and measurement noise. The filter predicts and updates using linearized matrices.
3.2 Application in Seismic Stations
Seismic stations in China traditionally rely on AC power with UPS, but the inherent randomness of earthquakes demands continuous power supply. Therefore, energy storage cells are essential. Many stations now use intelligent power controllers based on microprocessors like STM32F107, which provide display of temperature, humidity, battery voltage, and capacity, and prevent equipment damage by optimizing power logic. Remote power monitoring systems have been designed using TCP/IP protocols, allowing operators to remotely reset crashed instruments. Systems composed of power supply units, data acquisition cards, serial servers, and management software have changed maintenance patterns, improving efficiency and reducing costs. For example, in Fujian province, communication equipment in unattended stations was converted from AC to DC power supply, eliminating interference from DC-AC inverters and improving stability and data connectivity.
A typical remote power monitoring architecture includes: energy storage cell pack, charging controller, data acquisition module (voltage, current, temperature sensors), microcontroller, communication module (serial server or IoT gateway), and cloud/server-based management software. The system continuously logs energy storage cell parameters and sends alerts when anomalies are detected.
4. Conclusions
Online monitoring technology for energy storage cell status is vital for the safe and efficient operation of seismic stations. Real-time monitoring enables early detection of overheat, overcharge, overdischarge, and other hazards, preventing equipment shutdowns. It improves energy utilization by optimizing charging/discharging strategies based on accurate SOC. It extends battery life by identifying performance degradation trends and facilitating preventive maintenance. Moreover, it provides critical data for system-level optimization of the entire power infrastructure.
Future developments should focus on high-precision intelligent monitoring using advanced sensors and AI-driven algorithms to predict internal micro-states (e.g., cell-level temperature, voltage, impedance). Integration and miniaturization will allow installation in diverse applications without space constraints. Wireless communication and network-based platforms will enable remote centralized monitoring of large-scale energy storage cell arrays. Finally, deep integration with energy management systems will allow real-time optimization of energy distribution, creating a more intelligent and resilient power supply for seismic networks.
In summary, from a safety, reliability, and green economy perspective, designing an effective online monitoring system for energy storage cell packs—capable of real-time status display, parameter logging, and fault prediction—is a critical challenge and opportunity for the expansion and upgrade of national seismic monitoring networks. This work has profound theoretical significance for promoting the construction and application of modern seismic stations.