In this comprehensive review, we present a first-person perspective on the state-of-the-art online monitoring technology for energy storage battery systems, particularly focusing on their application in seismic stations. The reliability of energy storage batteries is critical for ensuring uninterrupted power supply in remote and unmanned seismic monitoring sites. Our work systematically examines the development of lithium-ion batteries, state‑of‑charge (SOC) estimation methods, equalization control strategies, and the current landscape of battery management systems (BMS) in both general and seismic station contexts. Through extensive analysis of recent literature, we aim to provide a clear understanding of the challenges and future directions in this field.
1. Introduction
Energy storage batteries serve as the core power source for seismic stations, especially those located in remote areas without reliable grid access. With the rapid expansion of national earthquake early warning networks, the number of stations equipped with battery backups has grown dramatically. For instance, the China Earthquake Early Warning Project has deployed over 15,000 stations, more than 5,000 of which rely on lead‑acid or lithium‑ion battery packs. Ensuring the safe, economical, and stable operation of these batteries is therefore paramount. Online monitoring technology enables real‑time assessment of battery parameters such as voltage, current, temperature, and state of charge (SOC), allowing predictive maintenance and fault prevention. This review summarizes the key advancements and open challenges in this domain, drawing upon recent research from both domestic and international sources.
2. Overview of Lithium‑ion Batteries for Energy Storage
Lithium‑ion batteries have become the dominant chemistry for modern energy storage due to their high energy density, long cycle life, low self‑discharge, and environmental friendliness. Table 1 summarizes the key advantages and disadvantages of lithium‑ion batteries compared to traditional lead‑acid batteries.
| Parameter | Lithium‑ion | Lead‑acid |
|---|---|---|
| Energy density (Wh/kg) | 100 – 270 | 30 – 50 |
| Nominal cell voltage (V) | 3.0 – 4.2 | 2.0 – 2.2 |
| Self‑discharge per month | < 1% | 3 – 20% |
| Cycle life (cycles) | 500 – 10000+ | 200 – 500 |
| Memory effect | No | Yes (if deep discharged) |
| Safety concerns | Thermal runaway possible | Gas generation, acid leaks |
| Cost | Higher | Lower |
Despite these advantages, lithium‑ion batteries require sophisticated management systems to monitor each cell’s state and prevent overcharge, over‑discharge, and thermal runaway. The evolution of lithium‑ion technology began in the 1970s with Whittingham’s first Li‑TiS₂ cell, progressed through the development of graphite anodes in the 1980s, and reached commercial maturity with Sony’s introduction of the LiCoO₂ cathode in 1992. Modern variants such as LiFePO₄ offer improved safety and longer life, making them ideal for stationary energy storage applications like seismic stations.
3. State‑of‑Charge (SOC) Estimation Methods
Accurate SOC estimation is fundamental to energy storage battery management. SOC indicates the remaining capacity as a percentage of the rated capacity. Over the years, researchers have developed numerous algorithms to improve estimation accuracy under dynamic operating conditions. We categorize these methods into three main families: (1) direct measurement (e.g., Coulomb counting), (2) model‑based approaches (e.g., Kalman filters), and (3) data‑driven approaches (e.g., neural networks). Table 2 provides a comparative summary of commonly used SOC estimation techniques.
| Method | Accuracy | Complexity | Robustness | Applicability |
|---|---|---|---|---|
| Coulomb counting | Low (accumulates error) | Very low | Low (requires reset) | Simple systems |
| Open‑circuit voltage (OCV) lookup | Moderate (requires rest) | Low | Moderate | Offline calibration |
| Extended Kalman filter (EKF) | High | Medium | High (handles noise) | Real‑time BMS |
| Unscented Kalman filter (UKF) | Higher than EKF | Medium‑High | High | Nonlinear systems |
| Particle filter | Very high | High | Very high | Heavy computation |
| Artificial neural network (ANN) | High | Medium (training offline) | Moderate (requires data) | Data‑driven |
| Hybrid (EKF + ANN) | Very high | High | Very high | Modern BMS |
Among model‑based methods, the extended Kalman filter (EKF) is widely adopted due to its balance between accuracy and computational efficiency. The discrete‑time state‑space model of a lithium‑ion cell, typically based on an equivalent circuit model (ECM), can be described as follows:
$$ \mathbf{x}_{k+1} = f(\mathbf{x}_k, u_k) + \mathbf{w}_k, \quad \mathbf{y}_k = h(\mathbf{x}_k, u_k) + \mathbf{v}_k $$
where \(\mathbf{x}_k = [SOC_k, V_{RC1,k}, V_{RC2,k}]^T\) is the state vector (including SOC and polarization voltages), \(u_k\) is the input current, \(f\) and \(h\) are nonlinear functions, and \(\mathbf{w}_k, \mathbf{v}_k\) are process and measurement noise, respectively. The EKF linearizes these functions using a first‑order Taylor expansion and recursively updates the state estimate and error covariance. A typical Thevenin ECM with one RC network yields the following differential equations:
$$ SOC_{k+1} = SOC_k – \frac{\eta \Delta t}{C_n} i_k, \quad V_{RC,k+1} = e^{-\Delta t / \tau} V_{RC,k} + R_1(1 – e^{-\Delta t / \tau}) i_k $$
where \(\eta\) is Coulombic efficiency, \(C_n\) the nominal capacity, \(\tau = R_1 C_1\) the time constant, and \(R_1, C_1\) the polarization resistance and capacitance. The terminal voltage is \(V_t = V_{OC}(SOC) – V_{RC} – R_0 i\).
Recent improvements include the use of adaptive EKF (AUKF) combined with BP neural networks, as proposed in several studies. For instance, a hybrid algorithm uses a trained BP network to provide a refined observation for the UKF, reducing estimation error to below 2% under dynamic loads. Another approach employs the sparrow search algorithm (SSA) to optimize the initial weights of an extreme learning machine (ELM), achieving faster convergence and higher accuracy compared to standard ELM.
4. Equalization Control for Energy Storage Battery Packs
In a series‑connected energy storage battery pack, cell‑to‑cell variations in capacity, impedance, and temperature lead to SOC imbalances over time. These imbalances reduce the usable capacity and accelerate aging. Equalization circuits and control strategies are therefore essential. We review three main categories: passive balancing (dissipative), active balancing (non‑dissipative), and hybrid approaches. Table 3 summarizes their characteristics.
| Type | Topology | Efficiency | Speed | Complexity |
|---|---|---|---|---|
| Passive (resistor shunt) | Resistor + switch per cell | Low (waste energy as heat) | Slow | Simple |
| Active (inductor‑based) | Buck‑boost, flyback | High (80‑95%) | Medium | Moderate |
| Active (capacitor‑based) | Switched capacitor / SC | Medium (70‑85%) | Fast | Moderate |
| Active (transformer‑based) | Multi‑winding transformer | Very high | Fast | High |
| Hybrid (active + passive) | e.g., BQ78PL116 + transformer | High | Fast | High |
One promising direction is the use of distributed control strategies for reconfigurable battery packs. In a two‑layer distributed system, each module communicates with its neighbors to achieve SOC consensus. The equalization step size can be computed based on voltage differences, reducing the convergence time. Another novel circuit uses a single pair of complementary square‑wave signals to control a switched‑capacitor network, significantly simplifying the gate driver requirements for large series strings. Hybrid methods combine the speed of active balancing with the low‑cost of passive resistors for fine adjustments.
5. Online Monitoring Systems for Energy Storage Batteries
Modern energy storage battery monitoring systems (BMS) integrate sensors, data acquisition, and communication modules to provide real‑time status. Table 4 lists representative BMS platforms and their key features from the literature.
| System Name | Developer | Key Features | Application |
|---|---|---|---|
| SmartGuard | Aerovironmevt (USA) | Distributed cell units, allows partial discharge while others charge via PV | Electric vehicles |
| BATTMAN | Hauck (Germany) | Software/hardware configurable for different battery chemistries | General BMS |
| Cloud BMS | TU Munich | Cloud‑connected, uses digital twin for SOH prediction | Automotive, grid storage |
| IoT‑based BMS | Texas A&M University | Raspberry Pi + Google Cloud, wireless data transmission | Large‑scale storage |
| Intelligent power controller | China Earthquake Administration | STM32F107 core, monitors battery voltage, temperature, capacity | Seismic stations |
Early systems, such as the PBWC system developed by EPRI in the 1980s, used fiber‑optic cables to transmit individual cell voltages and temperatures to a central monitor. Today, cloud‑based BMS platforms leverage IoT devices to collect data from each module and perform advanced analytics on the cloud. For example, the cloud BMS proposed by Baumann et al. (2018) uses an empirical aging model to predict remaining useful life and sends alerts when anomalies are detected. Kim et al. (2020) developed a small‑scale prototype using Raspberry Pi boards that communicate wirelessly, demonstrating that cloud resources can significantly enhance the accuracy of SOC and SOH estimation.
In the context of seismic stations, several specialized power controllers have been developed. The STM32F107‑based controller described in the literature provides local display of battery voltage, current, temperature, and estimated capacity. Another system uses a data acquisition card combined with a serial server to enable remote monitoring and control—allowing operators to reboot locked‑up equipment or switch power sources over TCP/IP. These systems have been deployed in unmanned stations in Fujian and other provinces, improving equipment uptime and reducing maintenance costs.
6. Specific Challenges and Solutions for Seismic Stations
Seismic stations present unique operational conditions for energy storage batteries: they are often located in remote, harsh environments with limited access; they must operate continuously for years with minimal human intervention; and they must withstand extreme temperatures, humidity, and seismic vibrations. These factors impose stringent requirements on the online monitoring system. We identify the following key challenges:
- Long‑term stability: Sensors and electronics must operate reliably for years without drift.
- Low power consumption: The monitoring system itself must draw minimal power to avoid draining the battery.
- Remote diagnosis: The system must support remote firmware updates and automated fault detection.
- High precision: SOC and SOH estimation must be accurate enough to schedule timely maintenance.
To address these, researchers have proposed adaptive algorithms that adjust the measurement noise covariance based on the operating current, improving SOC robustness under varying loads. For example, a modified EKF with current‑dependent noise covariance reduces estimation error by 50% compared to fixed‑noise EKF. Additionally, dual‑Kalman‑filter architectures have been used to simultaneously estimate SOC and capacity (SOH) in wind‑power smoothing applications, showing excellent performance even when battery parameters change over time.
7. Future Directions
Based on our review, we foresee several promising research directions for online monitoring of energy storage batteries:
- Digital twin integration: Creating a virtual replica of the physical battery that continuously learns from real‑time data and updates its models. This approach, already explored for automotive batteries, can be extended to stationary storage in seismic stations.
- Edge AI and federated learning: Moving some analytics to edge devices (e.g., microcontrollers) to reduce communication latency and preserve privacy, while sharing aggregated insights across a network of stations.
- Multiphysics modeling: Coupling electrochemical, thermal, and mechanical models to predict internal states (e.g., lithium plating, SEI growth) that lead to failure, enabling earlier warnings.
- Wireless, self‑powered sensors: Developing miniature sensors that harvest energy from the battery itself or from ambient vibrations to eliminate wired connections and reduce installation costs.
- Standardization and interoperability: Establishing common data formats and communication protocols for energy storage battery monitoring across different manufacturers and applications.
8. Conclusion
In this review, we have provided a comprehensive overview of online monitoring technology for energy storage batteries, with particular emphasis on their use in seismic stations. We discussed the evolution and properties of lithium‑ion batteries, compared various SOC estimation and equalization control methods, and surveyed existing monitoring systems. Our analysis shows that while significant progress has been made—especially in model‑based SOC estimation and active balancing circuits—challenges remain in terms of long‑term reliability, low‑power operation, and adaptability to diverse operating conditions. Future research should focus on integrating cloud computing, digital twins, and edge AI to create smarter, more resilient energy storage battery management systems. By ensuring the health and performance of these crucial components, we can guarantee the uninterrupted operation of seismic monitoring networks and contribute to the safety and resilience of our society.