GridSentinel — Self-Powered Grid Sensor Case Study | Abhay Kumar
Validated Concept · Phase 0

GridSentinel

A battery-free, clamp-on sensor that turns the electricity grid edge into a live stream of outage and fault data — harvesting its own power from the line it watches.

My role
Repositioning → spec
Status
v0.2 · Phase 0 bench
BOM per node
~CAD $95
Lifetime target
15+ years, no battery
The problem

Utilities find out about outages from phone calls

The distribution grid edge — the actual conductors carrying power to homes and businesses — is almost entirely unmonitored. When a section goes down, utilities typically learn about it from customer calls, not sensors, and crews patrol blind to find the fault.

  • →No plug, no accessible power source at most points on the line — any sensor needs to power itself.
  • →Storm restoration is slow and risky without knowing which spans are damaged or safe to re-energize.
  • →Electricity theft is a real cost, but the physics of proving it only close at certain points on the network.
Repositioning

The headline feature doesn't close the math on this line

Electricity-theft detection was the original pitch — but on a medium-voltage feeder, the energy-balance equation needed to prove theft doesn't close; it only works at the low-voltage transformer secondary. Rather than ship a headline feature that can't work where it was pitched, I re-scoped the first product around what a utility ops team actively wants and has no incentive to refuse: automatic outage detection and fault localization. Theft detection moves to a planned LV variant, where the physics actually work.

Discovery

Who has to say yes

OP

Utility operations · Primary

Wants faster outage detection and fault localization without new truck-rolls or infrastructure to maintain — the MVN's entire value case.

RP

Revenue protection · Later

Cares about theft detection — served by the LV/transformer-secondary variant once the MVN has proven itself in the field.

Backlog

User stories that shaped the wedge

US-01 As a utility ops manager, I want automatic section-level outage detection so I'm not relying entirely on customer calls. Must
US-02 As a utility ops manager, I want a fault localized to a specific span so crews drive to it instead of patrolling the whole feeder. Must
US-03 As a utility, I want the node to survive 15+ years with zero battery maintenance so the economics work across thousands of nodes. Must
US-04 As a storm-response team, I want damaged spans flagged as unsafe to re-energize so no one gets hurt during restoration. Should
US-05 As a revenue-protection team, I want an LV variant that closes the energy-balance equation so theft is actually detectable. Should
US-06 As a grid planner, I want dynamic line rating and power quality data from the same installed base so I get more value without new hardware. Could
Requirements

Spec'd as testable acceptance criteria

RequirementPriority
Harvest 100% of operating power from line current via split-core CT — no battery replacementMust
Detect de-energization and send a last-gasp message from stored energy before going darkMust
Localize a fault to the span between the last node that saw the surge and the first that didn'tMust
Hot-stick installable on an energized MV line in under 30 seconds, no outage requiredMust
Supercapacitor + primary-cell backup validated at 45°C+ ambient for the last-gasp budgetShould
Storm-mode tilt and arc sensing flags spans unsafe to re-energizeShould
LV / transformer-secondary variant with a resistor-divider voltage sense, enabling theft detectionCould
My role

Finding the wedge the physics actually support

I pressure-tested the original theft-detection pitch against the underlying electrical engineering, found it didn't close on a medium-voltage line, and re-sequenced the roadmap around a wedge utilities would adopt immediately. From there I specified the full hardware architecture, bill of materials, and phased validation plan.

What I designed

A node that powers, senses, and reports itself

Energy harvesting

Split-core CT → surge protection → harvest PMIC → supercapacitor + Li/SOCl₂ backup, feeding a regulated 3.3V rail with no plug.

Sub-GHz mesh comms

Wi-SUN self-healing mesh, pole-to-pole, backhauled to a gateway over cellular or fiber.

Self-orienting physical design

Clamps on as a gravity pendulum — self-orients vertically, keeps the antenna clear of the wire, hot-stick installable.

Detection logic

Outage, fault-localization, and storm-restoration logic, all running on the same MVN hardware — no separate device for each capability.

Under the hood

From current collapse to a crew dispatched to the right pole

1A fault drives a large current surge from the source side of the affected span
2Nodes upstream of the fault see the surge; nodes downstream don't
3Known radial topology localizes the fault to the exact span between them
4De-energized nodes send a last-gasp message from stored energy before going dark
5The mesh reports a section-level outage map — a crew drives to a span, not the whole feeder
Outcome

A de-risked platform, staged for a bench prototype

GridSentinel is fully specified through Phase 0 — architecture, BOM (~CAD $95/node electronics), and a de-risked product sequence that earns a place on the line with a wedge utilities want, rather than leading with the unsolved theft-detection problem. The dominant cost and risk, correctly identified upfront, is the custom HV enclosure and installation engineering — not the electronics.

Roadmap

Bench → pilot → platform

Phase 0

Bench prototype — prove harvesting, last-gasp, mesh, and fault-surge detection.

Phase 1

Single-feeder pilot — validate survival, outage detection, fault localization, and install time in the field.

Phase 2

LV theft-detection node and storm restoration, then arc/wildfire, dynamic line rating, and power quality as firmware on the installed base.

More case studies

See the rest of the work, or get in touch.