Abstract:The thermal runaway of ternary batteries involves electrochemical thermal mechanical multi field coupling effects. A single sensor can only capture local physical quantity changes and cannot reflect key factors such as gas yield and voltage fluctuations, resulting in increased early warning errors for thermal runaway. In response, a research on the design of an early warning system for thermal runaway of ternary lithium-ion batteries is proposed. In the design of system hardware units, it mainly includes sensor units, multi-channel data acquisition circuit units, main control chip units, and warning and execution structure units, forming a full chain hardware support system covering monitoring, acquisition, processing, and warning; In the system software module, first, real-time operation data of the ternary power lithium-ion battery is collected through negative temperature coefficient thermistor, fiber Bragg grating temperature sensor, infrared thermal imager, gas sensor, voltage/current sensor, pressure sensor, smoke sensor, and multi-channel data acquisition circuit, and its features are extracted and fused. Then, the extracted features are input into the time convolutional network risk prediction model for thermal runaway prediction, and finally, early warning layers are applied to batteries with thermal runaway risk. The experimental results show that the battery temperature distribution map obtained by the designed system is highly consistent with the actual battery temperature distribution map, and the minimum error in comprehensive feature vector extraction reaches 0.3%. The predicted probability of battery thermal runaway risk is consistent with the actual probability of battery thermal runaway risk, and the early warning results of battery thermal runaway are the same as the annotated results of the experimental samples.
