Abstract:Reliable and functional electric vehicle supply equipment (EVSE), which include electric vehicle (EV) chargers, are critical components in the global transition to EVs. Several studies have revealed that current EVSE reliability metrics, such as uptime, do not reflect the true reliability of EVSEs experienced by consumers. In this study, we have developed a novel tool that combines the powerful reconstruction capabilities of the Long Short Term Memory (LSTM) autoencoder with the long-term contextual awareness of an in-house naïve distribution method to learn the habitual usage patterns of EV chargers and effectively identify charging faults that may not be captured by traditional reliability measures.