Can you beat the model performance benchmark?

VIA seeks models to help address the most important clean energy challenges:

• 24-hour load prediction
• Anomaly detection
• NILM (non-intrusive load monitoring)

The challenge:

Three smart meter datasets are available:

• Low Carbon London dataset with 167 million rows (link)
• Reference Energy Disaggregation Dataset with 1.7 million rows (link)
• REFIT Electrical Load Measurements with 114 million rows (link)

Do you have a new idea for a model using one of the datasets? Can you improve on the existing models?

Utilities will be able to “upvote” and “downvote” a description of your proposed model to give you feedback before you begin model development. Interested utilities will reach out to prefered analysts through the secure VIA Insights Market™ (logins provided at the event) and be able to directly test your model on their data. Each utility’s data remains private and secure. You will receive a ranking of the performance of your model versus industry benchmarks.

Requirements:

Models must be in Tensorflow 2.0 protocol buffer (.pb) format and a link to your model must be submitted as a comment in your model proposal page on VIA Insights Market™.

Presentation

About VIA:

VIA is powering the clean energy revolution. VIA's AI platform helps organizations drive operational efficiencies and collaborate across the energy ecosystem by overcoming challenges related to data quality, data scarcity, and data privacy. We work with governments and the largest power providers in the US, Europe, and Asia. Now, we need your help to make an even bigger difference.