GridPulse AI

Predicting the future of energy to power a resilient tomorrow.

Comment

Inspiration I noticed a growing "predictability gap" in our global shift toward renewables. Unlike fossil fuels, wind and solar are inherently volatile, leaving grid operators buried in telemetry and the public confused by sudden price spikes or brownouts. I built GridPulse AI to act as a "Mission Control" for this era—a tool I designed to translate massive grid datasets into actionable intelligence for engineers and everyday residents alike.

What it does GridPulse AI is a real-time monitoring and predictive dashboard for electrical grids. I designed it to handle:

Live Monitoring: It tracks supply vs. demand in real-time for regions like Texas (ERCOT) and California (CAISO).

Physics-Based Simulation: I created models to simulate energy behavior for global cities like Tokyo and Cairo using urban density and weather patterns.

Stability Logic: I implemented a "15% Reserve Margin" rule; if the gap between supply and demand narrows too far, the system automatically triggers a "Caution" state.

Autonomous Resilience: In scenarios where demand exceeds supply, I programmed the system to simulate Virtual Power Plant (VPP) activation and execute Priority Load Shedding to protect hospitals and critical infrastructure.

Dual-Layer AI: I leveraged Gemini 3 Flash to generate two simultaneous outputs: a "Technical Brief" for experts and a "Simplified Summary" for the public.

How I built it I chose a high-performance stack to ensure the dashboard felt responsive and reliable:

Next.js 14 & Tailwind CSS: I developed a "Glow-UI" that shifts colors dynamically based on the real-time health of the grid.

Gemini 3 Flash: This serves as my analytical engine, processing telemetry every 5 seconds to generate advisories.

Recharts: I used this to visualize complex Gaussian distribution curves for "Morning Surges" and "Evening Peaks."

TypeScript: I wrote the underlying physics engine in TypeScript to calculate frequency drops and megawatt (MW) offsets with precision.

Challenges I ran into My biggest hurdle was Data Scarcity. High-resolution grid data is often proprietary or restricted for security reasons. To solve this, I built a Synthetic Physics-Based Simulation engine. This allowed me to fill the gaps so that even without a live API for a specific city, the AI’s behavior remained grounded in real electrical engineering principles, such as frequency stabilization.

Accomplishments that I'm proud of I’m most proud of creating the "AI Bridge." Usually, AI tools focus on either technical depth or simple accessibility. I successfully prompted Gemini 3 Flash to act as a "Bilingual Expert"—simultaneously outputting infrastructure jargon for engineers and plain English for residents without sacrificing accuracy.

What I learned I realized that Visibility = Resilience. By making the "15% Stability Rule" transparent to the user, the grid stops being a "black box." I also gained a much deeper understanding of how software, specifically through Virtual Power Plants (VPPs), can act as a decentralized battery to solve physical hardware shortages.

What's next for GridPulse AI My next goal is Local-Level VPP Integration. I want to move beyond simulation and allow users to connect their actual EV chargers and home batteries to the dashboard. By creating a literal "Pulse" of the city, I hope to help pave the way toward a 100% renewable grid that stays resilient.

Built With

Share this project:

Updates