Predictive Maintenance Solution for EV Charging Company

About the customer: Our partner specializes in modern EV charging solutions, offering certified charging stations, a reliable EV driver app, a charger management platform, and premium 24/7 support.

Business challenge: Like any device, EV chargers are vulnerable to sudden outages, causing inconvenience for end-users. To enhance customer experience and minimize downtime, the company sought to implement a predictive maintenance (PdM) solution to anticipate and prevent charger failures. 

Solution: Intelliarts analyzed historical EV charger data via the OCPP protocol to build an ML solution for predicting charger anomalies. Specifically, our data engineering team:

🔸 Identified ambiguity in health state labels and advised on automating and improving the labeling process.
🔸 Found gaps in diagnostic data, recommending ways to enhance data quality, including preserving raw data and using correct formats.
🔸 Developed an anomaly detection solution using DBSCAN, isolation forest, and local outlier factor algorithms to understand behaviors leading to outages.
🔸 Collaborated with technical experts to interpret findings, helping the company optimize charger performance.

Business outcome: As a result of this stage of the PdM project, our partner received detailed recommendations for enhancing their data collection pipeline to support future predictive model development. Specifically, we advised on gathering additional raw data, using the correct data format and types, automating data labeling, and setting up cold storage on AWS S3 for efficient long-term historical data storage.