New energy vehicles (NEVs) have become a fundamental part of transportation system. Performance of an NEV is hugely determined by batteries, motors, and embedded electric control units. In this paper, a comprehensive study that covers all these key components is presented. Mechanisms and characterizations of failures are given in detail. On top of these, algorithms for fault diagnosis are established based on big data of real-world NEVs with joint considerations of design flaws, usage behaviors, and environmental conditions. In this way, multiple types of faults can be detected ahead of time to avoid accident. Proposed methods have been verified by real-world operational data, indicating effectiveness while providing insights for NEV design optimization.