Read what the field is telling you.
Computer vision works through storm photos, aerial imagery, and written field reports to build a ranked view of what is likely damaged.
Storm restoration decision support for electric utilitiesAfter the storm, what gets fixed first?
Dovrane weighs the damage reports against your crews, your materials, road access, and who is on each feeder—then recommends a repair order, shows the alternatives it rejected, and explains why. Your operators make the call.
Working beta · Testing against simulated storms · Operators always make the call
14:21:04 UTC · EVENT TIME
Reports come in
Aerial photo E-184 and field ticket F-77 land in the event record.
01 / 06
PRODUCT BOUNDARY
Read-only advice that sits beside your OMS, GIS, ADMS, and work management.
- No autonomous dispatch
- No switching
- No grid control
Where the gap is
A damage map is not a restoration plan.
Your systems already tell you what broke, where it sits, and how the work gets tracked. None of them tell you what order to fix storm damage in. That gap—utility restoration sequencing—is the whole of what Dovrane does.
System you already runWhat it tells youWhat it does not
Weather and forecastingWhere the storm will hit and how many customers may go outWhich repairs to run first once the damage is real
Damage assessmentPhotos and field reports of what is brokenWhat to do when reports conflict or haven't come in yet
OMS and ADMSWho is out and what the network is doingWhich repair order restores the most customers, soonest
FLISR and automated restorationAutomatic isolation and reconfiguration for faults the scheme can clear on its ownThe repairs that still need a crew, and what order to send them in
GISWhere assets are and how they connectHow that interacts with crew skills, materials, and road access
Work managementHow approved work gets assigned and trackedWhether the order you chose was the right one
How the AI works
Read the damage. Weigh the limits. Compare the plans.
Dovrane applies computer vision and constrained optimization to one narrow problem in power grid restoration: the decision between knowing what broke and sending a crew to fix it.
Keep track of what you actually knew.
Every conclusion stays tied to the evidence behind it, when that evidence arrived, what was still missing, and where two reports disagreed.
Compare plans that could actually run.
The optimizer builds complete repair orders, scores them against your crews, materials, access, and topology, and throws out the ones that cannot be executed.
How it works, step by step
From the first damage report to a decision you can defend.
Each step records what you knew, what was still unclear, what limited the options, and who made the call.
- 01
What came in
Damage photos, field reports, weather, asset records, and read-only pulls from your existing systems—each stamped with the time it arrived.
- 02
What is likely broken
A ranked read on the damage, including where reports disagree, what hasn't been assessed, and where the model declines to guess.
- 03
What limits you
Feeder topology, crew classes and shift hours, materials on hand, road access, safety holds, and critical loads like hospitals.
- 04
Possible repair orders
Several complete restoration plans, each built from the same picture of the event at the same moment.
- 05
Reality check
Plans that break a known limit—no qualified crew, road closed, part unavailable—are held back rather than ranked.
- 06
The recommendation
One suggested repair order, with the estimated restoration curve, the runner-up plans, and why this one came out ahead.
- 07
Your call
Accept it, change it, hold it, or reject it. Your reason is saved alongside the recommendation.
- 08
What actually happened
Observed results are kept with the recommendation, so any decision can be replayed using only what was known at the time.
Every recommendation, fully explained
You should never have to take its word for it.
Dovrane shows the evidence, the tradeoffs, the plans it passed over, and the things it was not sure about. No black-box score.
- The evidence for it—and against it
- What limited the options, and which plans got cut
- The estimated restoration curve and customer-minutes out
- What the operator decided, why, and every version since
RECOMMENDATION / RR-0042
Checks passedIsolate S-17, then replace pole P-204
Est. customer-minutes out18,420modeled estimate
Customers back on2,180after step 02
Not sure aboutCrossarmneeds a field check
CUSTOMERS INTERRUPTEDEVENT TIME →
WHAT LIMITS THIS
No qualified crew free until 14:36 UTC
PLAN 03 — CUT
Road access to the site not confirmed
How we build it
Five rules we hold ourselves to.
We do one thing: help you choose the repair order, test that honestly, and leave the authority where it belongs.
It tells you when it does not know
Missing and conflicting reports stay visible instead of being smoothed over. “Not sure” is a valid answer.
Only plans that could actually run
Topology, crews, materials, road access, safety holds, and critical loads shape every option it offers.
You always see the tradeoffs
Each recommendation carries the alternatives, what limited them, and why this one ranked first.
Replay with only what you knew
Reviews rebuild the picture as it stood at the moment of the decision. No hindsight.
Measured against real outcomes
Repair orders are compared against your approved baselines and what actually happened on the ground.
Fits alongside what you already run
It reads. It never touches.
Dovrane takes a bounded, read-only feed from your systems and hands back a recommendation for your team to review. Nothing flows the other way.
OMS / ADMSGIS / topologyWork managementImagery / field reportsWeather / roadsCrews / materials
DOVRANEDecision intelligence
A recommendation for your teamEvidence · limits · alternatives · versions
No connection to dispatch, switching, or any operational technology.
How the read-only OMS and ADMS boundary worksProof before promises
We publish how we test before we publish results.
We report the full spread of results, the bad cases alongside the good, every constraint we broke, and the baseline we measured against—so you can judge the method, not just a number.
How we test itHOW WE TESTSimulated storms
What we measureEstimated customer-minutes out
What makes it fairSame information, same moment
ScenariosDevelopment · regression · sealed
What we reportFull spread · bad cases · violations
Every record hasEvidence · alternatives · limits
Questions we get
The things operators ask first.
Short answers to what comes up in the first conversation, including the ones where the answer is no.
- What order should we fix storm damage in?
- Dovrane recommends the repair order. It weighs incoming damage reports against your crews, materials, road access, feeder topology, and critical loads, then proposes a sequence, shows the plans it rejected, and explains why the top one ranked first. Your operators accept, change, hold, or reject the recommendation—Dovrane never dispatches a crew.
- How is this different from our OMS or ADMS?
- An OMS tells you who is out. An ADMS tells you what the network is doing. Neither tells you which repair to run first once the damage is real. Dovrane is a repair prioritization layer that sits beside them and reads from them—it does not replace either.
- Does it dispatch crews or perform switching?
- No. Dovrane is read-only and advisory. There is no control path from Dovrane to dispatch, switching, or any operational technology. Crew assignment, switching orders, clearances, and every real-world safety determination stay entirely with your utility.
- How do you measure whether a repair order was actually better?
- Dovrane measures a repair order primarily by estimated customer-minutes out (CMI): the number of customers off multiplied by how long they were off, summed across the event, compared against your approved baseline on the same event. Customer-minutes out is the raw quantity behind SAIDI, SAIFI, and CAIDI—we report it directly rather than as an index, because index definitions and exclusion rules vary by regulator. We also report the restoration curve, critical-load recovery, and how often the model broke a constraint or declined to answer.
- What data do you need from us?
- Dovrane needs a bounded, read-only feed: damage reports and imagery, asset and topology context, crew and material availability, and road status. Nothing flows back the other way. For a demo you do not need to send anything at all—we run it on a simulated storm.
- Do you have utility deployments or published results?
- No. Dovrane is a working beta testing against simulated storms. There are no utility deployments and no validated performance results, and we will not imply otherwise. We publish the method we use to test before we publish any number that comes out of it.
- Does Dovrane produce ETRs?
- No. Dovrane recommends a repair order and shows the modeled restoration curve behind it. Estimated times of restoration (ETRs) published to customers stay with your OMS and your communications process.
- How does this handle mutual aid crews?
- Mutual aid crews enter as additional crew capacity with their own qualifications, shift hours, and arrival times. Because they usually do not know the territory, road access and site reachability weigh more heavily in the plans that assign them.
- Is this power outage restoration software?
- Not in the usual sense. Outage management software tells you who is out and tracks restoration as it happens. Dovrane does grid restoration optimization: it takes the damage reports, crews, materials, road access, and feeder topology and recommends what order to repair in, then records why. It is advisory and read-only, and it sits beside your OMS rather than replacing it.
Next step
Bring us your hardest storm.
We would rather show you the method than pitch you a number. Start a conversation about a bounded look at one of your past events.