The emergence of distributed renewable power generation capacity across the urban landscape presents a new challenge to power utilities. Micro-generation creates transient pools of capacity that must find available load to consume it. The traditional grid infrastructure is to lossy, to complicated, to focused on centralized generation, in a word, it's inadequate.
Micro capacity is best consumed as close as possible to where it's generated. What if utilities could send the load to the capacity instead of distributing the capacity on a grid? In a few years, cities will be filled with roving autonomous electric vehicles periodically looking for the cheapest recharge opportunity. We propose to harness this mobile "load" to dynamically match localized capacity by predicting micro-capacity and guiding (via price incentives) recharge seeking vehicles to the optimal charging station.
Historical weather and generation data was used to train a predictive model to provide predicted local capacity based on current weather conditions. Weather and available charge station (parking spaces are used as a stand-in) are retrieved from CityIQ. Additional environmental input is dynamically detected through a Predix Edge device. We built a micro-service that, on demand, retrieves the weather (temperature) and charge station availability, runs the trained models and accumulates a map of predicted capacity. The application displays this visually, and responds to requests from roving vehicles with charging opportunities in a given timeframe.
Challenges: CityIQ is limited in its weather condition reporting and doesn't currently monitor charging infrastructure. Integration with ESRI geospatial services was harder than expected.
Accomplishments: we pulled together a diverse set of resources to create trained models and working services in almost real time.
What's next: we'd like to add a geospatial hierarchical aspect (keep zooming out until enough matchable load and capacity are found) and add additional prediction factors.