Fault diagnosis for inconsistent cells in a battery pack is important for electric vehicle (EV) safety. Discovering the fault and identifying its cause on-site are extremely challenging. The literature has rarely explored such issues. This paper presents a battery pack system in a demonstrated EV with 96 cells in series and discovers the battery power fade fault during the demonstration. The preliminary analysis, after collecting data and preprocessing the typical data periods, shows that the internal or contact resistance increase causes the fault. To further identify the fault cause, we introduce an equivalent circuit model (ECM) and identify the cell resistances using the total least squares algorithm. A diagnostic method is proposed, using the identified cell resistances from 3 months demonstration data. Calculating the Shannon entropy clearly identifies the cause of the power fade fault. Appropriate measures are taken to solve the fault, and the latent safety issue is eliminated. The proposed diagnostic method can be used on-site for EV diagnosis without laboratory testing methods. (C) 2012 Elsevier B.V. All rights reserved.