Abstract: With the large-scale application of electrochemical energy storage, thermal runaway detection and timely warning research of lithium battery is of great significance for ensuring the safe operation of energy storage power station. In this paper, a test platform for the thermal runaway performance of lithium battery is set up. The sound signal of blade energy storage lithium battery under thermal runaway condition is tested. The acoustic characteristics and change rules of different development stages of thermal runaway such as battery bulging, pressure relief valve opening, venting, explosion and fire are analyzed. It shows the effectiveness of sound feature for thermal runaway detection. Aiming at the problem that the sound characteristics of thermal runaway are easily disturbed by the noise of other equipment such as ventilation fans and PCS in the energy storage battery compartment, an anti-interference analysis method of thermal runaway sound signal based on wavelet packet decomposition and spectral subtraction speech enhancement is proposed. The analysis result shows that the proposed method can effectively eliminate the wind noise interference from the thermal runaway sound signal, and can accurately restore the battery thermal runaway sound characteristics. After denoising, the correlation coefficient between the original and the denoised sound signals reaches 0.96. The results can provide as technical reference for thermal runaway detection and early warning of the energy storage lithium battery.

Key words: energy storage power station, blade energy storage lithium batteries, thermal runaway, sound feature, anti-interference analysis, acoustic detection, wavelet packet, spectral subtraction