AI Can Now Predict Storm Power Outages Hours Ahead

Power outages from storms cost time, money, and sometimes safety. If utility companies knew exactly which neighborhoods might lose power hours in advance, they could send repair crews early and warn you personally. That’s why researchers in Texas tested whether modern AI chatbots — the same kind of technology behind ChatGPT — can predict outage risk just by reading weather reports.

They used six years of real outage records and high-resolution weather data from central Texas. The AI was asked to judge, three, six, and twelve hours ahead, whether severe weather would likely knock out power. The catch: the chatbots were given zero examples of past outages — they had to figure it out from general knowledge. Their results were compared to traditional machine learning models that were specially trained on the Texas data.

The old-style models were better at correctly identifying which outages really happened, and they made fewer false alarms. But the newest AI chatbots scored surprisingly close — good enough to be useful. What makes them special is their flexibility. They can explain their reasoning in plain language, telling engineers why a risk is high. And they can instantly apply their logic to any region, without needing years of local data to train on. In contrast, traditional models need large amounts of historical data and can struggle in new territories.

The researchers suggest the best idea isn’t to pick one over the other. Instead, utilities could use trained machine learning for accuracy and use AI chatbots to check results, explain risks, or help predict outages in areas where data is scarce. For you, that could mean more reliable early warnings, fewer surprise blackouts, and faster power restoration after bad weather.