Building the Power Grid Dashboard
The Hormuz Strait crisis, oil price spike and rumors of an increase in power blackouts got me wondering: Are power grids actually at risk, or is the media just running with "crisis" headlines? I wanted to find out for myself, so I built yet another dashboard.
The Global Power Grid Dashboard pulls open data going back to 2000 and runs a linear regression to show whether outages are actually increasing over time. Spoiler: they are, and the trend is accelerating. But the reasons are more nuanced than "oil bad."
The build has a few things I hadn't done before, and I do love leaning on Claude in these situations:
Statistical projection from raw data
The dashboard runs a simple linear regression on 26 years of outage data and plots a trend line with confidence bands. If the current year's annualized pace exceeds the band, something structural might be changing. Writing the projection model from scratch instead of using a library forced me to actually understand the math.
Multi-metric chart toggling
One chart, three views: event count, customers affected, MW lost. Each tells a different story. Event count shows frequency is climbing. Customers affected shows the damage is spikier. MW lost shows the scale of cascading failures.
Cause-shift analysis
Side-by-side doughnut charts comparing 2000-2012 vs 2013-present. Weather-related outages have roughly tripled. Cyber incidents went from zero to double digits. Equipment failures are steady. Demand-driven outages are flat. The grid isn't failing because of one thing. The threat mix is shifting.
Regional grid cards with honest data-quality flags
Every region gets a card, but most of the world has terrible open grid data. I marked cards with "Limited Data" or "No Open Data" instead of pretending the numbers are reliable. Only North America and Europe have real transparency.
Oil price overlay
Brent crude charted alongside grid events, with key crisis dates annotated. The insight the dashboard surfaces: oil crises are economic crises for OECD nations, not blackout crises. The actual blackout impact lands on countries that were already fuel-import-fragile like Bangladesh, Pakistan, Ghana.
EIA live data integration
The dashboard can pull real-time US generation and demand data from the EIA Open Data API. This was the first time Claude and I worked with their API. Clean data, good documentation.
The look
The visual style is warm amber on dark, kind of a control room aesthetic with film grain and scanline overlays. IBM Plex Sans for readability, JetBrains Mono for data. It works well for a data-heavy page without feeling clinical.
Biggest learning
The gap between "the news says X is a crisis" and "the data says Y is actually happening" is massive. Building the dashboard didn't just surface information. It changed how I read the headlines. I'm not done looking into this!