Fault diagnosis for lithium-ion battery energy storage systems based on local outlier factor

Lithium-ion batteries (LIBs), when faulty or operating under abnormal conditions, can cause fire accidents, consequently, the enhancement of LIBs safety is a key priority for their large-scale application. This goal can be achieved by fault diagnosis, which aims detecting the abuse conditions and diagnosing the faulty batteries at the early stage to prevent them from developing into thermal runaway. In this work, the local outlier factor (LOF) method is adopted to conduct fault diagnosis for energy storage systems based on LIBs (LIB ESSs). Two input generation algorithms, i.e., the multiple factors at single time step input generation (MFST) algorithm and the single factor at multiple time steps input generation (SFMT) algorithm are proposed for the LOF method. Moreover, in order to simulate different severe levels of internal short circuit (ISC), an ISC model is added to the electrical-thermal coupled model for an air-cooled LIB ESS. Then the performance of the LOF method in detecting different severe levels of ISC are studied based on the simulated data from this air-cooled LIB ESS as well as the experimental data from a water-cooled LIB ESS. The LOF method is proved to be effective in detecting the faulty cell at three different ISC severe levels (with 1 omega, 10 omega and 100 omega ISC resistance, respectively) in the air-cooled LIB ESS and two faulty cells in which the equivalent ISC resistances are 100 omega and 25 omega, respectively, in the water-cooled LIB ESS.