This paper introduces a data-driven anomaly detection framework designed to enhance the safety and reliability of lithium-ion battery packs deployed in large-scale electric transport systems. Leveraging principal component analysis (PCA) and cumulative sum (CUSUM) control charts, the method monitors mean-based residuals of voltage and temperature across cell groups to detect early-stage faults such as internal short circuits, sensor failures, and thermal irregularities. Experimental validation using real-world data from a battery-electric locomotive demonstrates the system’s ability to identify anomalies with deviations as low as 4mV and 0.15°C while maintaining a falsepositive rate below 3%. The approach reduces detection time by 56% and missed anomalies by 60% compared to conventional thresholding methods. By integrating real-time fault diagnosis into battery management systems, this work contributes directly to the advancement of safe, durable, and sustainable battery manufacturing and energy storage deployment in heavy-duty electric mobility.