Random Forest-Based Online Detection and Location of Internal Short Circuits in Lithium Battery Energy Storage Systems With Limited Number of Sensors

For fault detection in energy storage systems, the current topologies and detection methods require a large number of sensors. Therefore, this article proposes a random forest (RF)-based online detection and localization method to monitor faulty cells in lithium battery energy storage systems. First, the internal short circuit (ISC) is diagnosed by combining voltage and current signals to reduce the number of sensors used for monitoring. The feature extraction method proposed in this article requires at most $2M+1$ sensors in the $N \times M$ topology, which can reduce the use of $M(N-2)-1$ sensors compared to current research methods. Then, using RF to process the extracted features reduces the detection and localization time. Finally, constant-current discharge 100- $\Omega $ ISC resistance data are used to train the RF model, and different ISC resistance data for New European Driving Cycle (NEDC) conditions are used to verify the accuracy and precision of the trained RF model. The results show that the topology optimization and algorithm proposed in this article can use simple sample data of constant-current discharge to train the algorithm and monitor fault information and location in milliseconds during practical applications.