Lithium-ion batteries can reach a safety-critical state when operating outside normal conditions. In the worst case, a number of (strongly) exothermic side reactions result in the thermal runaway of the cell. In this case, a large amount of energy and potentially toxic reaction products are emitted immediately. The operating parameters are currently monitored by the BMS so that a safe operation is generally guaranteed. However, safety-critical conditions, which can be caused by external impacts, ageing under unfavourable conditions or manufacturing defects, cannot be completely excluded, so that in practice serious accidents with lithium-ion batteries occur frequently. For this reason, it is of great interest to detect these safety-critical conditions in advance in order to minimize the risk and consequences of a battery failure. A possible approach is described in more detail in the following.
The modelling of the behaviour of lithium-ion batteries by a equivalent circuit model has proved successful many times due to the quite simple parameterisation and the possible online implementation caused by the rather low computing time. This model approach will be used to calculate the electrical and thermal behaviour of a cell for a dynamic load. In the case of an failure, thermal, exothermic side reactions will occur in the cell. The cell will be heated more than the simulation calculates for the given load. The early failure detection is now based on comparing the voltage calculated by the model with the measured voltage. If there is an unrealistic difference between the two voltage values, the system generates a warning signal. The measured and calculated temperature is also considered for optimization and model verification.
This approach has several advantages for practical implementation compared to conventional temperature monitoring. Because of the sometimes very transient heat distribution when failures occur, the measured temperature at the cell housing is significantly lower than inside the cell. Thus, temperature monitoring is delayed and furthermore only effective after the exceeding of the temperature limit. The voltage behavior, in contrast, is directly influenced by the temperature in the whole cell and the malfunctions can be detected before the temperature exceeds the operating limit. In addition, the reduction of temperature sensors in the battery system is also interesting for cost reasons.