Project Report: VoltSync

Track Chosen

Smart Grid

VoltSync aligns with the Smart Grid track because it focuses on real-time demand-supply balancing, distributed energy resource (DER) coordination, and AI-driven load management for utilities and microgrids during peak stress events.


Problem Statement Addressed

Distributed energy resources — rooftop solar, home batteries, EVs, smart thermostats — are growing faster than utilities can coordinate them. During peak load events (heat waves, cold snaps, solar ramp-downs at dusk), grid operators face a widening gap:

  • They know demand is spiking but can't see or control the distributed assets behind the meter
  • Demand response programs exist, but dispatching them is manual, slow, and one-size-fits-all
  • Battery and EV capacity sits idle because there's no real-time signal telling owners or aggregators when to discharge
  • Curtailment decisions (rolling blackouts, load shedding) happen too late and too bluntly, hitting vulnerable communities hardest

Today, most utilities coordinate DERs through static schedules, day-ahead signals, or manual overrides. The result is wasted flexibility on calm days and chaotic scrambling on critical ones.


Ideation and Development Process

We started by asking: what would it take for a utility operator to treat thousands of distributed assets as a single, controllable virtual power plant — without needing a PhD in optimization or a six-month integration project?

Our first pass was a simple "discharge all batteries at peak" trigger, but we quickly realized the value is in sequencing and prioritization — which assets, in what order, for how long, without violating owner preferences or grid constraints.

We iterated around four principles:

  1. Forecast-first orchestration: Use load and generation forecasts (demand curve + solar/wind predictions) to identify stress windows before they arrive, not after frequency drops.
  2. Asset-aware dispatch: Every DER has different capacity, state-of-charge, owner constraints, and grid location. Dispatch plans must respect all of these, not blast a uniform signal.
  3. Equity-aware shedding: When curtailment is unavoidable, the system should protect vulnerable zones and essential services first, not default to whoever is easiest to disconnect.
  4. Operator-grade control: Outputs must be dispatchable task lists and SCADA-compatible setpoints, not just dashboards and charts.

We scoped the build around an end-to-end loop: forecast → stress detection → DER dispatch plan → curtailment fallback → post-event learning.


Final Solution and Intended Impact

VoltSync is an AI demand response copilot for utilities and microgrid operators.

It turns real-time load and generation forecasts into a coordinated DER dispatch plan that maximizes available distributed flexibility before resorting to curtailment — personalized to each operator's fleet, constraints, and equity priorities.

Core capabilities

  • Peak stress prediction: Ingests load forecasts, solar/wind generation curves, and grid capacity data to flag stress windows 2–48 hours ahead with confidence-ranked severity.
  • DER fleet visibility: Aggregates available capacity across batteries, EVs, smart thermostats, and controllable loads — broken down by zone, state-of-charge, and owner-set constraints.
  • Optimal dispatch sequencing: Generates a time-phased dispatch plan that sequences which assets to activate, in what order, and for how long — maximizing peak shaving while respecting device limits and owner preferences.
  • Equity-weighted curtailment: When dispatch alone isn't enough, produces a curtailment plan that protects hospitals, shelters, and vulnerable communities first, with an adjustable tradeoff between "minimize total unserved energy" and "protect critical loads."
  • Supermemory personalization: Learns from uploaded fleet data, operating procedures, rate structures, and past event logs to tailor dispatch strategies to each operator's real constraints and vocabulary.
  • Exportable outputs: Produces a control room dispatch brief, aggregator-ready setpoint schedules, and a public-facing conservation alert for customer communication.
  • Post-event learning loop: Logs actual asset response vs. planned dispatch, flags underperforming assets or constraint violations, and refines models for the next event.

Intended impact

VoltSync helps operators unlock flexibility they already have but can't coordinate:

  • Fewer and shorter curtailment events by exhausting distributed flexibility first
  • Faster, more precise demand response that respects device-level and owner-level constraints
  • Equitable load shedding when curtailment is unavoidable
  • Continuous improvement across peak events instead of rebuilding dispatch logic from scratch each time

In short, VoltSync upgrades peak load management from blunt curtailment to intelligent, asset-aware, equity-first orchestration.

MLH track: Gemini API and Auth0

Built With

  • custom-reasoning-models-(planner-+-scorer-+-verifier)
  • fastapi
  • gemini-api
  • javascript
  • nws/baron-weather-api
  • python
  • react/next.js
  • supermemory-(rag)
Share this project:

Updates