Abstract: The detection of the internal short circuit (ISC) faults in a li-ion battery energy storage system is restricted by the real-time performance of the online detection and the availability of the monitoring data. This is an urgent problem to be solved in the safe operation of the li-ion battery energy storage system. This paper proposes an detection method of the ISC faults in a li-ion battery energy storage system based on the CUSUM event segment identification and the improved spectral clustering algorithm (SCA). Firstly, considering the voltage/temperature characteristics of the ISC fault, the suspected ISC fault event segment is identified based on the CUSUM transient event detection algorithm. Secondly, the 3D fault features are constructed to characterize the feature attributes of the short-circuit faults in the detection object. Then, the characteristic distance matrix of the ISC fault based on the Wasserstein measure is constructed to detect the sparsity characteristics of the points in the three-dimensional space, and objectively delineate the fault clustering in order to achieve the detection of the ISC fault. The experiment platform of the li-ion battery is built, and the electric-thermal coupling simulation model of the li-ion battery is established based on the measured data. The results show that the proposed method is able to accurately identify the suspected ISC fault event segment, and realize the fault detection among different series, parallel forms and fault types, which proves the correctness and feasibility of the proposed method.