1. Introduction
Energy storage battery, particularly lithium iron phosphate (LFP) batteries, have become pivotal in addressing the intermittency and volatility of renewable energy sources like wind and solar power. Despite their advantages in stability, safety, and cost-effectiveness, LFP-based energy storage systems face critical challenges, such as thermal runaway caused by overcharging, which poses significant safety risks. Existing monitoring methods—such as voltage, temperature, and gas detection—rely on external parameters and suffer from delayed responses to internal battery degradation. This work introduces an ultrasonic detection (UT)-based early warning system to monitor internal state changes in energy storage battery, enabling timely intervention before catastrophic failure.
2. Methodology
2.1 Experimental Setup
We designed a comprehensive testing platform for LFP energy storage battery, integrating ultrasonic transmission, temperature, voltage, current, surface deformation, and gas concentration measurements. The system comprises:
- Ultrasonic transmission subsystem: A Gaussian-modulated sinusoidal pulse generator, amplifier, and oscilloscope.
- Thermal runaway characterization subsystem: A temperature-controlled chamber, charge-discharge cycler, and gas/strain sensors.
Battery samples (20 Ah aluminum-cased LFP cells) were preconditioned with three 0.5 C charge-discharge cycles to ensure consistency.
2.2 Ultrasonic Signal Analysis
Key ultrasonic parameters include:
- Signal intensity (Vrmsrms): Calculated as:Vrms=1N∑i=1NVi2Vrms=N1i=1∑NVi2where ViVi is the voltage of the ii-th data point, and NN is the total number of points.
- Time of flight (TOF): The duration for ultrasonic waves to traverse the battery.
Normalized parameters were defined as:RV=Vrms(t)Vrms(0),RT=TOF(t)TOF(0)RV=Vrms(0)Vrms(t),RT=TOF(0)TOF(t)
3. Experimental Results
3.1 Static State Analysis
3.1.1 Temperature Effects
Temperature significantly impacts ultrasonic signals (Table 1):
- RV decreases with rising temperature due to reduced electrolyte impedance mismatch.
- RT increases due to thermal expansion and decreased ultrasonic wave velocity.
| Temperature (°C) | RV (Normalized) | RT (Normalized) |
|---|---|---|
| 20 | 1.00 | 1.00 |
| 35 | 0.65 | 1.05 |
| 45 | 0.20 | 1.10 |
A sharp inflection occurs at 35°C, attributed to phase transitions in the electrolyte.
3.1.2 State of Charge (SOC) Effects
SOC influences electrode elasticity and lithium-ion distribution:
- RV increases by 11.56% as SOC rises from 0% to 100%.
- RT decreases by 0.67% due to enhanced wave velocity in graphite anodes.
3.2 Dynamic Charge-Discharge Behavior
Under cyclic conditions, ultrasonic signals correlate with temperature and SOC (Table 2).
| Charging Rate (C) | RV Variation (%) | RT Variation (%) | Dominant Factor |
|---|---|---|---|
| 0.25 | 8.2 | 1.5 | Temperature |
| 0.50 | 9.1 | 2.1 | Temperature |
| 0.75 | 10.3 | 3.0 | Temperature |
Key observations:
- RV better reflects SOC variations, while RT is temperature-dependent.
- Polarization effects cause abrupt RV changes at cycle start/end.
3.3 Overcharge-Induced Thermal Runaway
Overcharging triggers irreversible internal changes (Figure 1):
- Stage I: Voltage rises due to lithium plating and SEI layer degradation.
- Stage II: Gas generation (H22, CO22, CO) increases internal pressure, causing surface deformation.
- Stage III: Ultrasonic signals attenuate completely (RV ≈ 0, TOF indeterminate) before venting valve activation.
| Charging Rate (C) | Signal Attenuation Time (s) | Venting Valve Activation (s) |
|---|---|---|
| 0.25 | 1150 | 5154 |
| 0.50 | 586 | 2418 |
| 0.75 | 257 | 1537 |
Key Insight: Ultrasonic signal decay precedes gas release and temperature spikes by 30 minutes, enabling early warnings.
4. Early Warning Algorithm
We propose an Improved Mahalanobis-Taguchi System (MTS) to assess battery health using six parameters: voltage, temperature, RV, RT, H22 concentration, and surface strain.
4.1 Algorithm Workflow
- Data Preprocessing: Z-score normalization and feature weighting via Fisher’s criterion.
- Baseline Space Construction: Healthy-state data define reference Mahalanobis distances.
- Anomaly Detection: Weighted Mahalanobis distance (WMD) thresholds trigger alerts.
Thresholds derived from 3σ准则:Threshold=μWMD+3σWMDThreshold=μWMD+3σWMD
where μWMD=0.9998μWMD=0.9998, σWMD=0.97σWMD=0.97.
4.2 Validation
Testing on a 0.5 C overcharged energy storage battery demonstrated:
- HI (Health Index) dropped below threshold at 7262 s (SOC = 104.2%).
- Warning time: 30 minutes earlier than venting valve activation.
5. Conclusion
- RV is more sensitive to temperature and SOC variations than RT, making it ideal for dynamic monitoring of energy storage battery.
- Ultrasonic signal attenuation serves as a critical early indicator of overcharge-induced thermal runaway.
- The MTS-based algorithm enables 30-minute advance warnings, significantly enhancing the safety of LFP energy storage systems.
This work bridges the gap between internal state detection and external parameter monitoring, offering a robust framework for safeguarding energy storage battery against catastrophic failures.