Inspiration
Extreme weather cuts power supply at the exact moment heating or cooling demand spikes. An outage map alone never explains why one substation keeps a hospital online while another drops critical loads. GridShield turns power-grid stress testing into an interactive, inspectable engineering workbench that anyone can verify locally.
What it does
- Linearized DC Power-Flow and Islanding Solver. Choose one of three synthetic community grid scenarios (Winter Freeze Emergency, Summer Heat-Dome Peak, or Hardened Microgrid Baseline) and any hour of the day. Open any transmission line or trip any generator to watch partial-pivot Gaussian elimination re-solve flows across every energized island, balancing generation and served demand without rounding branch ratings.
- Full N-1 Contingency Sweep and Cascading Protection Trips. Automatically tests every surviving line and generator individually, tracks modeled thermal protection trips, and separates the highest branch loading from the worst loss of critical service so isolated hospitals are never hidden behind low line loading.
- 24-Hour Resilience Investment Comparator and Auditable Exports. Compare battery storage (with charge/discharge losses and reserve floor), gas weatherization, residential demand response, and hospital microgrid islanding against the baseline over 24 hours. Exports formula-neutralized CSV schedules, ranked N-1 outage tables, a SHA-256 canonical JSON engineering record, and a clean one-page printable resilience brief.
How we built it
GridShield runs offline in plain ES-module JavaScript, HTML5, CSS3, and inline SVG with zero runtime dependencies and zero external network calls. Partial-pivot Gaussian elimination solves B * theta = P per energized island, disconnected intermittent renewables are never counted as grid-forming sources, and Web Crypto generates canonical SHA-256 digests over every exported audit bundle.
Challenges we ran into
The most heavily loaded surviving branch is not always the contingency that hurts people most. An isolated critical bus can go dark with zero overloaded lines remaining. Separating worst branch loading from worst critical-load outage, validating island connectivity after every cascade step, and hardening CSV exports against C1 control-character formula injection required rigorous fuzzing across 7,000 synthetic grids.
Accomplishments that we're proud of
- On the synthetic Summer Heat-Dome Peak scenario at 6 p.m. (48 MWh / 14 MW battery, weatherized gas, residential demand response, hospital islanding off), daily unserved energy drops from 574.66 MWh to 409.91 MWh (164.75 MWh avoided), critical-load uptime rises from 70.83% to 91.67%, and selected-hour worst N-1 loading falls by 28.55 percentage points while honestly reporting that one branch still exceeds its rating.
- Passed 283 automated Node tests and 45 headless Chrome checks at 1280x800 and 390x844 with zero console errors, zero external requests, and one-page printing, after closing all findings from a three-lane adversarial audit covering physics, security, and WCAG contrast.
What we learned
Higher served energy does not automatically guarantee N-1 contingency safety. Presenting both metrics side by side teaches communities and students how storage, weatherization, demand response, and microgrid islanding solve different failure modes.
What's next for GridShield
Add a documented benchmark against a standard IEEE public test case and explore dynamic thermal line ratings with look-ahead battery dispatch. GridShield is an educational synthetic workbench and not an operating aid for a live utility.
Built With
- css3
- html5
- javascript
- node.js
- power-systems
- svg
- web-crypto-api