SolarPanelWatcher

Instead of flying drones over solar panel array to detect issues, I wanted to automate the process and us AI and IoT to detect issues with the panels and send a notification to the repair team.

Comment

Inspiration

  • Its difficult to tell when a solar panel has an issue, where that issue occurred and what caused the issue.

What it does

  • The solution uses AI to detect heat anomalies in a solar panel array and notifies the repair crew.

How we built it

  • I used images of a circuit breaker panel taken with a seek thermal imager.
  • I trained an AI vision module in Customvision.ai
  • I downloaded into an Azure Percept Dev Kit
  • I fed the images of the breaker panel into device with a cellphone held up to the Percept Camera
  • I routed the messages through Azure Stream Analytics to blob storage, CosmosDB, and live into PowerBI.

Challenges we ran into

  • Installation of IoT Edge Runtime on a Raspberry Pi
  • Parsing JSON arrays for PowerBI live connection
  • Arducam commands on raspberry pi to capture thermal image
  • customvision.ai debug of image rotations
  • timezone conversion of date time stamp
  • setting up IoT Central connection parameters

Accomplishments that we're proud of

  • trying out CosmosDB connection to PowerBI for the first time.
  • Applying query language from a different project to this project's JSON messages.
  • Making great progress on learning about installing IoT edge runtime on a Raspberry Pi
  • Connection to parallel paths in Stream Analytics....ie, putting data into PowerBI and Blob Storage at the same time.
  • Training/Feeding Azure Percept with images from a thermal imager as a simulation.
  • Just getting all the different pieces of the puzzle to talk to each other and the moment that live data was going from the thermal image on a cell phone through the Percept and all the way into a PowerBI report!!!

What we learned

  • Cosmos DB - NoSQL...wow....
  • Stream analytics
  • PowerBI live connections
  • Connecting Cosmos and Blob and live streams to PowerBI
  • Watching live streams, and data in stream analytics coming from the Percept
  • Learned about Logic Apps and trigger sounds and emails from stream analytics.
  • Visual Studio code to watch messages on IoT hub
  • Data explorer to view Blob and CosmosDB data

What's next for SolarPanelWatcher

  • Contact a solar panel array installer and test on the real panels
  • Add microphone and noise data to Percept model
  • add recording of images using Percept Camera after heat issue detected.
  • apply to other things that give off heat when there is an issue

Built With

  • azure
  • blobstorage
  • cosmosdb
  • iothub
  • percept
  • powerbi
  • streamanalytics
Share this project:

Updates