Safety: The algorithm needs to be able to have accurate data to prevent unwanted behavior
Reliability: The algorithm should perform in various complex scenarios
Ease of Operation: The algorithm and model should be able to be deployed by a power company even if they don't understand machine learning
Minimum maintenance: The algorithm should operate autonomously without interference from the power company
Compatibility: The algorithm should work on any smart grid