Predictive Maintenance Solution for EV Charging Company

After years of successful collaboration on software solution development, our partner that offers modern electric vehicle (EV) charging solutions with premium 24/7 customer service reached out to us with one more business challenge. The company aims at creating the best possible customer experience, and that’s why they asked us to help to build a predictive maintenance (PdM) solution to minimize downtime and maintain charging stations in top condition.

Solution

We have conducted a detailed analysis of historical data received from EV chargers via OCPP protocol and provided our partner with important insights on the existing data quality and the additional data that was missing to build the PdM solution we planned for. We also completed a data anomaly analysis, which helped us understand which anomaly behavior usually results in the EV charger outages.

Here is a brief summary of the current state of the project:

• We began our research with the initial analysis of historical data received from EV chargers via the OCPP protocol. The insights we’ve got showed the ambiguity in the existing EV charging station health state labeling, so we helped the customer to improve and automate the labeling process based on the latter.
• We researched ways to improve the state diagnosis data quality and predictive power by preserving the raw historical data instead of aggregated one and by using the proper format and type of data.
• The anomaly detection analysis also helped us discover the charging sessions with a really long duration. This observation didn’t look normal to us, so we decided to invest more effort in this investigation and, hence, provided the customer with deeper insights into their equipment behavior.
• To find anomalies, our data scientist team used three unsupervised anomaly detection algorithms, namely DBSCAN, the isolation forest algorithm, and the local outlier factor. As a result, we found two data clusters with normal behavior and one cluster with abnormal behavior. Together with the station technical experts, we made an effort to interpret this finding from the business perspective, so the company could improve the behavior of their EV charging stations based on this.

Results

The data science team provided our partner with a set of specific recommendations as for the additional raw data to be collected and the proper format and type to use to provide predictive power to ML models. Intelliarts also advised on ways of improving the customer’s data collection pipeline, including the automation of data labeling process and building proper cold storage on top of AWS S3 for effective long-term storing of historical data. 

Now our software development engineers are working side by side with the customer to implement those recommendations and are upgrading their EV Charging stations to the next generation with the increased number of sensors. As soon as the customer collects enough amount of EV charger diagnosis data in the suggested format, we will continue our collaboration on building the ML-powered PdM solution.