Last year, as Hurricane Helene carved its way across the Southeast, transmission and distribution operators were reminded what “storm duty” really means. The grid doesn’t just creak under strain — it bends and threatens to snap. Demand spikes, circuits overload, field crews are scattered, and communication systems falter just when we need them most.
This is the moment when vendors say automation, AI, and advanced forecasting will save us. But when you’re staring at a wall of alarms at 2 a.m., running on your third energy drink, the question isn’t whether the tools are flashy. The question is simple: when the grid is breaking, can we actually trust automation?
The Operator’s Reality in Extreme Weather
Storms are nothing like the calm, air-conditioned command center you see when you take the tour of a control center on a Tuesday afternoon. You know the one where you look around at all the operators and wonder quietly, if you're smart, why it is they're being paid to sit around. It means 16 on 8 off shifts, stacked back-to-back, staring at outages and alarms rolling in like an endless waterfall, trying to decide which constraint matters most. It means listening to three radio channels, while your phone rings off the hook, all while leadership stops by with a chipper inquisition trying to determine the extent of the damage when you really have no idea just yet.
The truth: operators make life-or-death decisions while running on fumes. Fatigue, cognitive overload, and sheer stress blur the line between confidence and doubt. Every click carries risk and every delay compounds it.
The Promise of AI, Forecasting, and DERMS
To be fair, the technology landscape looks promising on paper. Weather-driven load forecasts claim to predict the next peak. DERMS and ADMS platforms promise automated feeder switching, FLISR schemes, and crew dispatch recommendations. AI algorithms churn out ranked options: “Isolate here. Shed this feeder. Deploy this crew.”
In theory, that’s a lifeline — faster pattern recognition, fewer blind spots, and better clarity when the storm hits. For Control Center leadership who are constantly trying to drive down outage response times, and eliminate errors, the draw is real while to the exhausted operator, the pitch is seductive: automation as a second set of eyes that never blinks.
But anyone who’s spent time in a control room knows — theory and reality rarely align once the grid starts to break.
The Trust Gap: Where Automation Breaks Down
Trust erodes fast when conditions degrade. Forecasting models work until the storm drifts 40 miles off track or ice builds thicker than predicted. A DERMS tool can only optimize if the data feeding it is clean — and during a storm, “garbage in, garbage out” becomes the rule, not the exception. Anyone who has worked a storm can tell you that sometimes the hardest part is simply keeping track of your configuration.
Worse, over-automation can create a black-box effect. The tool says to open a breaker, but it can’t explain why. In clear skies, that might fly. In chaos, it doesn’t.
When telemetry is missing, comms are spotty, and alarms flood the EMS — algorithms freeze, and the operator is left holding the bag.
That’s why resilience starts with trust — and trust begins with design choices that center the operator, not the algorithm.
Operator-Centric Resilience: Turning Lessons into Action
Automation isn’t the enemy — bad integration is. The goal isn’t to replace operators; it’s to make technology feel like a teammate, not a distraction.
Here’s how to get there.
1. Baseline & Audit — Conduct a Control-Room Automation Audit
Why it matters: You can’t move forward without understanding where you are now. Most utilities don’t have a clear inventory of which tools exist, how they’re used, or where they fail.
How to implement: Correlate data latency across systems, map every automation dependency, and survey operators about which tools they actually trust — and which they quietly ignore.
2. Data Quality First — Fix the Foundation
Why it matters: “Garbage in, garbage out” isn’t just a cliché, it’s why automation collapses in storms. Data blind spots, failed telemetry, and stale SCADA points poison even the smartest tools.
How to implement: Audit telemetry reliability, flag stale feeds, and create a bad-data playbook so operators know which sources to distrust when things degrade., and how to get them fixed.
3. Co-Design with Operators — Build Trust Early
Why it matters: Tools built without operator input usually fail under stress. Operators know what decisions actually feel like at 3 a.m. when comms are down.
How to implement: Involve control-room staff in vendor design sessions. Run workshops where operators critique workflows before rollout. Reward candid feedback, not compliance.
4. Make the AI Explain Itself — Don’t Trust Black Boxes
Why it matters: If a system can’t explain why it’s recommending an action, operators won’t execute it when stakes are high. Transparency is the foundation of trust.
How to implement: Require reason and confidence scores for every recommendation. Build an “Explain Mode” toggle that shows the assumptions and data behind each decision.
5. Train for Chaos — Not Perfection
Why it matters: Most training assumes clean data and rested operators — but storms are the opposite.
How to implement: Inject failures into drills. Simulate telemetry loss, conflicting alarms, and time pressure. Let operators practice with half-truths, not perfect visibility.
6. Deploy Incrementally — Earn Trust One Storm at a Time
Why it matters: Operator trust isn’t built in demos, it’s earned in the field.
How to implement: Start with low-risk automation (crew routing, feeder prioritization). Run in a sandbox during minor events to compare algorithm output against human decisions. Come back after the event and run a post-mortem to see where the automation and the operators can get better.
7. Build Guardrails — Keep the Human in the Loop
Why it matters: Automation should never remove operator control. The moment someone feels locked out, trust evaporates.
How to implement: Define override and rollback protocols. Make “manual mode” accessible in two clicks. After each event, verify the override path worked as expected.
8. Learn After Every Event — Institutionalize Feedback
Why it matters: Every storm reveals blind spots in data, design, or training. Most are forgotten by the next week.
How to implement: Make post-event reviews about more than restoration time. Capture what automation got wrong, what operators ignored, and why. Feed that data into retraining and design updates.
9. Lead with Accountability and Not Blame
Why it matters: When automation makes a bad call, finger-pointing kills progress. Accountability builds resilience.
How to implement: Define ownership for automation decisions and escalation paths. Make it safe to challenge the tool, even mid-event.
Looking Ahead
Extreme weather isn’t slowing down. Polar vortexes, heat domes, and late-season hurricanes are now constants, not anomalies. Utilities can’t afford to ignore automation — but they can’t afford blind trust either.
AI and DERMS must be deployed with operators, not around them. Trust is built when tools are transparent, explainable, and human-centered.
When the storm comes, and it will, the systems that endure will be the ones that extend human judgment, not replace it. Because you can’t automate judgment; you can only build tools that make good judgment faster.
In the control room, trust isn’t given to the smartest algorithm. It’s earned, one decision and one storm at a time.