Tesla PowerWall ML Predictions
Tesla PowerWall ML Predictions
Embedded Files
The goal is to ingest historical electricity usage, along with weather data, and predict power consumption for tomorrow. Merge prediction with utility plan rate data to determine when to discharge PowerWall for maximum savings while maintaining maximum backup reserves
Ingest historical weather data for local area (temperature, hour by hour) - python
Download historical power consumption from Tesla Gateway (python API)
Merge and group data into hour-by-hour consumption (kWh) and temperature (python)
Load data into Google Vertex AI - Tabular prediction
Deploy trained ML model to Google Cloud Function
Schedule day-ahead prediction by downloading predicted weather, invoking Google Cloud Function API (python)
Adjust time-of-day pricing and reserve capacity thresholds for PowerWall (python API)
End result: Personal (random) power consumption created a lot of unpredictable consumption patterns. For example, charging the car up before a road trip, or being away for a few days significantly impact energy usage.
With less than two years of data, it turned out there were far too few data points to make any meaningful predictions.
Page updated
Report abuse