Remote DCFC Reliability and Downtime Detection Tool: Detecting EV Charging Failures That Standard Reliability Protocols Cannot Detect

The Caltrans ZEV 30-30 project is established to support California's goal of deploying five million ZEVs on the road by 2030, focusing on filling gaps in the State Highway System's corridor ZEV charging network. The reliability of DC Fast Charging (DCFC) infrastructure along these corridors is essential. Unreliable chargers erode driver confidence and undermine the transition to zero-emission transportation. Charging Station Operators (CSOs) currently rely on conventional monitoring protocols, primarily the Open Charge Point Protocol (OCPP), to detect charger failures. While OCPP-based monitoring is effective for detecting most electrical and software failures, it cannot identify a broad class of faults arising from mechanical damage, physical obstruction, network communication outages, or logistical barriers. [1, 2] These hidden issues persist until an EV driver encounters the faulty charger and reports the problem, leading to delayed fault resolution and degraded consumer experience.This report presents a predictive anomaly detection tool developed under Caltrans Agreement No. 65A1188 that enables Charging Station Operators to detect hidden charging faults by analyzing habitual EV driver usage patterns. The tool incorporates two anomaly detection models: a naïve probability distribution-based technique and a Long Short-Term Memory (LSTM) autoencoder.