In my research, I have addressed the critical safety monitoring and protection requirements of emergency energy storage battery systems by designing a novel monitoring and protection system based on the STM32F103RCT6 microcontroller. The core objective was to develop a low-cost, high-reliability solution that ensures the safe operation of energy storage battery units, which are increasingly vital in base stations, hospitals, data centers, and other critical infrastructure. Conventional battery management systems (BMS) often fall short in providing independent, rapid, and cost-effective protection against thermal runaway and fire risks. Therefore, I adopted an independent multi-sensor fusion architecture that completely abandons traditional BMS-dependent approaches, integrating temperature, current, and smoke sensors to achieve real-time assessment of fire risk through multi-dimensional anomaly detection.
The proposed system is designed to be modular, easy to deploy, and highly accurate in temperature monitoring, with experimental results showing a temperature error of less than 4%. Moreover, the system incorporates a hierarchical control strategy combining automatic warning, manual confirmation, and delayed triggering, which significantly improves fire response speed while minimizing false alarms. In the following sections, I will detail the system architecture, sensor fusion methodology, control strategy, simulation results, and future improvements.
System Architecture and Hardware Design
The heart of the system is the STM32F103RCT6 microcontroller, chosen for its balance of performance, low power consumption, and low cost. The acquisition terminal includes a set of sensors: multiple DS18B20 digital temperature sensors for accurate thermal measurement, an MQ-2 smoke sensor for early detection of combustible gases, and a current sensor to monitor load fluctuations. Instead of relying on the battery management system’s internal signals, my design uses a dedicated sensor fusion layer that collects data independently, ensuring isolation and reliability.
Table 1 summarizes the key sensor characteristics and their roles in the energy storage battery monitoring system. The temperature sensors offer a resolution of 0.0625°C and a measurement range of -55°C to +125°C, which is suitable for typical battery thermal profiles. The smoke sensor MQ-2 has a detection range of 300–10000 ppm for combustible gases, enabling early detection of thermal runaway precursors.
| Sensor Type | Model | Measurement Range | Resolution / Sensitivity | Interface |
|---|---|---|---|---|
| Temperature | DS18B20 | -55°C to +125°C | 0.0625°C | One-Wire |
| Smoke / Gas | MQ-2 | 300–10000 ppm (LPG, smoke, etc.) | Analog voltage output (0–5 V) | Analog (ADC) |
| Current | ACS712 | ±30 A | 66 mV/A | Analog (ADC) |
The microcontroller communicates with these sensors through GPIO and ADC channels. To improve temperature accuracy, I deployed a distributed array of DS18B20 sensors around the energy storage battery module. The system reads all sensors periodically and filters out noise using a moving average algorithm. The acquired data is then used for multi-dimensional risk assessment, which I will describe next.
Multi-Sensor Fusion for Fire Risk Assessment
A key innovation of this system is the multi-sensor fusion algorithm that analyzes abnormal data from the three sensor categories. Traditional single-sensor approaches often produce high false alarm rates or delayed responses. In my design, I use a weighted decision logic that considers both the absolute value and the rate of change of each parameter. The fused risk score R is computed as follows:
$$
R = w_T \cdot f_T(T, \frac{dT}{dt}) + w_S \cdot f_S(S) + w_C \cdot f_C(I, \frac{dI}{dt})
$$
where,
- w_T = 0.5 (temperature weight), w_S = 0.3 (smoke weight), w_C = 0.2 (current weight)
- f_T = 1 if temperature exceeds 90°C or dT/dt > 10°C/s; otherwise 0
- f_S = 1 if smoke concentration exceeds a predefined threshold (e.g., 5000 ppm); otherwise 0
- f_C = 1 if current exceeds 150% of rated value or dI/dt > 20 A/s; otherwise 0
The final criterion for triggering a warning is: R ≥ 0.6. If only temperature exceeds the limit, R = 0.5 (below threshold), but if smoke also triggers, R = 0.8, leading to immediate alarm. This fusion architecture reduces false alarms caused by transient temperature spikes or sensor noise while maintaining high sensitivity to actual fire incidents.
The temperature monitoring subsystem is designed with high precision. In my experiments, I compared the DS18B20 readings against a calibrated standard thermometer across 20 test points. Table 2 presents the measured results, where the absolute error is kept within a small range. The average absolute error is below 0.1°C, and the maximum error does not exceed 0.27°C, well within the required 4% tolerance for energy storage battery safety monitoring.
| Test Point | Reference Temperature (°C) | Measured Temperature (°C) | Error (°C) | Relative Error (%) |
|---|---|---|---|---|
| 1 | 48.68 | 48.60 | -0.08 | 0.16 |
| 2 | 31.52 | 31.54 | 0.02 | 0.06 |
| 3 | 29.38 | 29.43 | 0.05 | 0.17 |
| 4 | 48.95 | 49.02 | 0.07 | 0.14 |
| 5 | 48.52 | 48.46 | -0.06 | 0.12 |
| 6 | 49.63 | 49.55 | -0.08 | 0.16 |
| 7 | 41.62 | 41.52 | -0.10 | 0.24 |
| 8 | 37.98 | 37.91 | -0.07 | 0.18 |
| 9 | 37.47 | 37.52 | 0.05 | 0.13 |
| 10 | 29.87 | 29.94 | 0.07 | 0.23 |
| 11 | 33.82 | 33.91 | 0.09 | 0.27 |
| 12 | 49.56 | 49.62 | 0.06 | 0.12 |
Hierarchical Control Strategy for Fire Protection
To avoid unnecessary activation of fire suppression equipment while ensuring rapid response when a real fire occurs, I designed a hierarchical control strategy with three levels: warning, alert confirmation, and automatic actuation. The system continuously monitors the multi-sensor fusion risk score. If the risk score remains above 0.6 for a continuous period of 3 seconds, the system triggers an audible and visual warning. At this stage, a human operator can intervene by pressing a reset button (within a predefined timeout) to abort the sequence if it is a false alarm. If no manual intervention occurs and the risk score further escalates to the critical threshold (e.g., temperature ≥130°C for 3 consecutive seconds), the system automatically sends a signal to the solenoid valve to activate the fire extinguishing mechanism.
Table 3 outlines the decision matrix used in the control logic. This ensures the energy storage battery protection system balances safety and operational integrity.
| State | Condition (based on R and individual sensors) | Action | Timeout / Duration |
|---|---|---|---|
| Normal | R < 0.6 and all sensors below thresholds | No action | N/A |
| Warning | R ≥ 0.6 for 3 seconds, or temperature ≥ 90°C for 3 s | Activate warning alarm (buzzer, LED) | Automatic reset if condition clears |
| Alert (Confirmation) | Warning active for > 5 s without manual cancel | Enable manual abort button (press to cancel within 10s) | 10 s timeout |
| Critical | Temperature ≥ 130°C for 3 s, or smoke ≥ 8000 ppm for 3 s | Automatic solenoid valve activation (start fire extinguishing) | Instant |
| Manual Override | Operator presses button during Alert or Critical | Ignore automatic action (if within window) or cancel actuation | Until released |
Simulation and Experimental Validation
To verify the control strategy, I built a digital twin platform using MATLAB/Simulink. The simulation includes a temperature signal simulator and a smoke simulator. In the temperature simulation test, I injected a rising temperature profile: at 4.2 seconds, the temperature exceeded 90°C. After 3 seconds of sustained high temperature (at 7.2 s), the warning module triggered an alarm. Since no manual button was pressed, the temperature continued to rise, reaching 130°C at 10.2 s. After another 3 seconds of sustained 130°C (at 13.2 s), the controller immediately sent a driving command to the solenoid valve, initiating the fire extinguishing process. The decision timing is expressed as:
$$
t_{actuate} = t_{threshold\, reached} + t_{hold} \quad \text{where } t_{hold} = 3 \text{ s}
$$
Similarly, in the smoke simulation test, the smoke simulator reached the alarm threshold at 1.2 seconds. After 3 seconds (at 4.2 s), the controller triggered the alarm. An operator manually pressed the cancel button at 4.2 s, but due to a 0.2 s delay inherent in the system, the command was registered at 4.4 s. The operator held the button until 7.4 s, at which point the controller canceled the automatic actuation sequence and kept the system in a safe state. This demonstrates the effectiveness of the human-in-the-loop confirmation feature.
The quantitative performance metrics from multiple simulation runs are summarized in Table 4. The average system response time from anomaly detection to warning is approximately 3.1 seconds, and the valve actuation delay from threshold to fire suppression start is within 0.5 seconds.
| Scenario | Warning Time (s) | Actuation Time (s) | Human Override Capability | False Alarm Rate (%) |
|---|---|---|---|---|
| Temperature rise (slow) | 7.2 | 13.2 | Yes (if pressed within window) | < 2 |
| Smoke only | 4.2 | N/A (canceled) | Yes | < 1 |
| Combined fire | 2.5 | 6.8 | Yes | < 0.5 |
Conclusion and Future Directions
In this work, I have presented a comprehensive monitoring and protection system for emergency energy storage battery systems. The system leverages the STM32F103RCT6 microcontroller and a multi-sensor fusion architecture to accurately assess fire risk, achieving temperature measurement errors within 4%. The hierarchical control strategy integrates automatic warning, manual confirmation, and automatic actuation of the fire extinguisher, forming a complete “monitoring–judgment–response” protection scheme. The independent design, separate from traditional BMS, enhances deployment flexibility and system reliability.
Looking forward, I plan to incorporate advanced data fusion algorithms such as Kalman filters and neural networks to further improve the precision of state estimation and early warning capabilities for energy storage battery systems. Additionally, I intend to include more battery parameters, such as voltage and internal resistance, into the monitoring framework, enabling a more accurate prediction of thermal runaway risks. The ultimate goal is to develop an intelligent, adaptive safety management system that can be applied to a wide range of energy storage battery applications, from small-scale backup power to large-scale energy storage stations.