The Plug In Sensor That Watches the Whole Grid A Fire Prevention Gadget Turned Into a National Power Quality Instrument

This application covers a network of small sensors plugged into ordinary household power outlets, reporting the shape of the mains voltage waveform to a central server that compares one house against its own history and against its neighbours to spot surges, sags, brownouts, frequency wander, loose neutral connections and generators starting up. It was filed by Whisker Labs, Inc., the American company behind the Ting home electrical fire sensor, and names Christopher Dale Sloop, Robert S. Marshall, Donnie Bixler and Chonglin Liu as inventors. The application is a divisional of Australian application 2021233035 and claims priority from United States provisional application 62/989,415 filed on 13 March 2020.

The Problem

The background section builds its case out of public numbers rather than assertion. Citing the US Energy Information Administration, it notes that the average American electricity customer went without power for 250 minutes across 1.3 outages in 2016, and that the figure nearly doubled in 2017 to 470 minutes, or 7.8 hours, across 1.4 outages. The longest outages went from the order of 20 hours in 2016 to a little over 40 hours in 2017. The 2017 figures match the EIA’s own published summary of its 2017 reliability data, which reports 7.8 hours and 1.4 interruptions for the average customer and notes that most of the increase came from hurricanes and winter storms. Many of these outages are unplanned, and the specification points out the obvious consequence that a householder who is away from home does not know one has happened.

Detecting an outage is harder than it sounds, because the detector loses power at the same moment the house does. The specification notes that existing outage detection devices rely on a backup battery or generator, which adds cost, has a finite life, creates a maintenance obligation, and can introduce a delay between the outage starting and the backup supply coming up.

The second half of the background is about power quality rather than outright loss of supply, and it is the part the claims actually pursue. Generation is becoming a mixture of wind, solar, nuclear, battery, gas and coal, with on-premise generation added on top, all superimposed on a grid of varying age. Switching between generation types causes surges and sags. Ageing transformers and connections deteriorate and fail. Surges, sags and failing equipment damage appliances and create conditions for electrocution and fire, and residential electrical fires are particularly lethal because they start inside walls and cavities where nobody sees them. The document puts the annual American toll at 420 deaths, 1,370 injuries and 1.4 billion dollars in residential damage.

Then it turns the argument outward. A homeowner notices lights flickering without being told there is a serious problem in the connection to their house. Utilities, meanwhile, carry increasing liability for what their own deteriorating plant does. The specification cites a Pacific Gas and Electric fire incident data report recording over 2,400 grid-caused fires between 2014 and 2019, more than 13 billion dollars in liability and the company’s bankruptcy filing; a Texas Wildfire Mitigation Project study finding 4,000 fires caused by transmission or distribution events in under four years; transformer explosions in Madison, Wisconsin and at an American Electric Power substation in Texas in 2019; and the February 2021 Texas grid failure. The gap it identifies is that nobody is measuring the low voltage end of the network, in the homes, at a resolution and density that would show these problems developing.

What This Invention Does

Before describing claim 1, one thing has to be said plainly, because the title and the abstract will mislead a casual reader. The title is “Methods and systems for detection and notification of power outages and power quality”, and the abstract describes an outage detection scheme in detail: the sensor sends periodic keepalive packets, watches for a rising edge in each clock cycle of a nine millisecond timer, fires a fault packet when the edge arrives early or not at all, and the server declares an outage when the keepalives then stop. None of that is claimed here. All thirty-two claims of this divisional are directed to power quality. The words keepalive and fault packet do not appear in the claim set at all. Outage appears only inside the generator detection claims, as a precondition. If you want the keepalive invention, it is in the parent, not in this document.

Claim 1 is a system claim with two halves. The first half is one or more sensor devices, each coupled to a circuit, each configured to detect an input signal generated by electrical activity on that circuit, generate an output signal from it, and transmit power quality data based on the output signal to a server. The second half is the server, configured to receive that power quality data, analyse it together with historical power quality data from the same sensors to detect one or more power quality events, and transmit a power quality notification to remote computing devices based on the detected events.

Two words in that claim carry the weight. “Historical” means the system is not comparing the voltage against a fixed rulebook but against what this particular house normally looks like. And “one or more sensor devices” is what makes claim 3 possible, which adds the ability to correlate a detected event at one sensor with events at other sensors, and with external events entirely. Claim 15 says those external events include lightning activity, electrical grid monitoring events and energy pricing events.

The hardware is unassuming. The body describes an opto-isolator connected across the hot and neutral wires with a ground reference, monitoring a 120 volt 60 hertz supply, feeding a processor and an analogue to digital converter. It names the device: “An exemplary power outage detection device 102 is the Ting Sensor available from Whisker Labs, Inc. of Germantown, Maryland.” Zero crossings on a healthy 60 hertz supply arrive about 8.33 milliseconds apart. The device’s timer counts down from 9 milliseconds, which the specification labels as about 55 hertz, and a rising edge that arrives sooner than 7.6 milliseconds, about 65 hertz, or that does not arrive before the timer expires, is treated as a loss of power. For power quality the sensor sends five RMS voltage samples every quarter second, or 20 readings a second, which is enough resolution to place an event on the ITI, formerly CBEMA, curve of voltage amplitude against duration. The body gives the intuition for that curve with a worked pair: 300 volts for about 16 milliseconds generally harms nothing, while 300 volts held for more than five seconds can damage anything on the network.

The dependent claims are where the system stops being generic. Claim 2 lists the event vocabulary: surge, surge jump, sag, sag jump, brownout, swell jump, high frequency filter jump, frequency jump, recurring power quality problems, phase angle jump, loose neutral and generator activation. Claims 4 through 14 then define how several of them are detected, and the body attaches example numbers. A surge is generated when RMS voltage exceeds a threshold percentage of nominal, given as 120 per cent, for a number of consecutive data points, with claim 7 adding that the threshold varies with how many consecutive points are involved, which is the CBEMA curve expressed as a rule. A brownout uses the same structure below nominal, with 70 per cent as the example, over a queue of about six seconds of data. Correlated events are those within a defined time window, given as 400 milliseconds, and a defined proximity, given as ten kilometres, which is how a set of individual surges becomes a single “grid surge event”.

Two of the detections are genuinely clever. The loose neutral test in claim 4 works from the absence of correlation rather than its presence: the server looks at surges, surge jumps and sags recorded by a single sensor that do not match events from any nearby sensor, and calls a loose neutral when the daily averages cross thresholds, with the body giving more than one surge per day, more than ten surge jumps per day exceeding 10 per cent of nominal voltage, or more than ten sags per day, measured over the previous seven days. A loose neutral is a genuinely dangerous fault in a split phase house, and the specification explains its signature: with the neutral disconnected, the two legs are no longer balanced and the voltage on one leg jumps upward as appliances on the other leg switch on and off. If the same pattern shows up across several homes on one transformer instead of in one home alone, the specification reads it as a failing transformer or substation rather than a fault inside a house.

The generator detection in claim 14 is the other one. A single sensor records an outage, and then within a defined period, given as sixty seconds, the standard deviation of the mains frequency jumps above a threshold given as 0.05 hertz, and that change does not match any correlated external event. The system calls that a “Generator On” event, because a small engine driven generator holds frequency far less tightly than the grid does. When the standard deviation later drops below 0.025 hertz, it declares “Generator Off”.

Notifications go two ways. The body describes sending surge and loose neutral notifications to a remote device belonging to the householder, and also to a remote device belonging to the utility or grid operator serving the home, so that they can act on a fault they did not detect themselves.

Key Features

• Comparison against the home’s own history. Claim 1 requires the server to analyse incoming power quality data in conjunction with historical data from the same sensors, so an event is a departure from that house’s normal, not from a fixed standard.
• Correlation across neighbouring sensors. Claim 3 allows an event at one sensor to be correlated with events at other sensors and with external events, using an example window of 400 milliseconds and a proximity of ten kilometres to promote individual readings into a grid event.
• Loose neutral detected by what does not correlate. Claim 4 identifies a dangerous in-home wiring fault precisely because its surges and sags appear at one sensor and nowhere nearby, with example rates of one surge, ten large surge jumps or ten sags per day over a week.
• A duration dependent voltage threshold. Claims 7 and 9 vary the surge and brownout threshold percentages according to how many consecutive readings are involved, which encodes the ITI voltage tolerance curve rather than a single trip point.
• Generator activation inferred from frequency jitter. Claim 14 detects a backup generator starting by looking for a frequency standard deviation above roughly 0.05 hertz within about sixty seconds of an outage, with no matching external event.
• Twenty voltage readings a second from a wall socket. The body specifies five RMS samples every quarter second, obtained by an opto-isolator across hot and neutral in a plug-in device, with no connection to the switchboard or meter.
• Notification to the utility as well as the householder. The body sends the same detected events to a remote computing device associated with the utility serving the home, framing the consumer device as a source of grid intelligence.

Who Is Behind It

Whisker Labs, Inc. of Germantown, Maryland is named on the title page, and the specification names its product outright at paragraph 63, the Ting Sensor. That is unusually direct for a patent specification and it removes any doubt about what the claimed sensor device is.

The corporate name has since moved. The whiskerlabs.com domain now redirects to tingfire.com, and the site’s footer carries a Ting Labs, Inc. copyright at the same Germantown address, so readers should treat Whisker Labs and Ting Labs as the same business under a new consumer-facing name rather than assume the applicant has disappeared. The company’s own impact page claims more than one million homes protected, more than 30,000 fires prevented, and grid visibility covering 98 per cent of American neighbourhoods, and describes selling that visibility to utilities, insurers, emergency responders and news organisations. The sensor is sold at 99 dollars or supplied free through more than thirty insurance carriers, which is how a network of that size gets built without asking a utility to pay for meters.

The inventor list is the interesting part. Christopher Dale Sloop and Robert S. Marshall are the company’s Chief Technology Officer and Chief Executive Officer respectively, and both are listed as co-founders on the company’s team page. Both also appear in the founder list of Automated Weather Source, the Germantown, Maryland company founded in 1992 that built and ran the school-based weather station network behind WeatherBug and later became Earth Networks. Marshall and Sloop came to that partnership as engineers from EAI Corporation. The pattern is the same one twice: take a cheap sensor, put a very large number of them in ordinary buildings, and sell the aggregated view of a phenomenon that was previously only measurable by the institution that owned the infrastructure. They did it with weather, and this specification is them doing it with the distribution grid.

On priority, the specification is explicit rather than silent. The Related Applications paragraph claims priority to United States provisional application 62/989,415 filed 13 March 2020, and the title page records the document as a divisional of Australian application 2021233035. That makes the priority country the United States.

Why It Matters

Utilities generally find out about an outage when a customer telephones them, or when a smart meter stops answering, and they find out about deteriorating power quality when something fails. Distribution networks are instrumented well at the substation and thinly after that. The final transformer, the service drop and the customer’s own switchboard are close to invisible, which is precisely where most of the faults that start fires live.

What a consumer device network changes is the density and the ownership of the measurement. A hundred thousand sensors inside houses on one distribution network see the low voltage side continuously, from inside the customer premises, at twenty readings a second. That produces a view no utility has of its own network, and the correlation logic in claim 3 is what converts it from a hundred thousand separate household anomalies into a map of which transformer is failing. The specification’s own framing, that a recurring problem in one home is likely a loose neutral while the same problem across homes on one transformer indicates a failure at the transformer, is the whole argument in a sentence.

There are two honest caveats. The first is that the specification’s numbers are almost all given as examples: 120 per cent, 70 per cent, ten kilometres, 400 milliseconds, 0.05 hertz. Claim 1 itself contains no numbers at all, which makes it broad and correspondingly vulnerable, and the numeric substance sits in dependent claims that are easier to design around. The second is that the whole system depends on a commercial sensor fleet that exists because consumers want fire protection. The grid instrument is a by-product of a home safety product, and it lasts only as long as people keep buying and plugging in the sensor. That is a genuinely different basis for critical infrastructure monitoring than a regulated utility asset, and it is worth understanding that the data now being cited in wildfire investigations comes from it.

Related Concepts

• Electric power quality – the general measure of how far a supply’s voltage, frequency and waveform depart from ideal, which is what claim 1 detects.
• Voltage sag – the short duration drop in RMS voltage that claim 2 lists among the detectable event types.
• Brownout (electricity) – the extended undervoltage condition the system flags when RMS voltage falls below about 70 per cent of nominal.
• Root mean square – the averaging method behind the RMS voltage readings the sensor sends twenty times a second.
• Power outage – the loss of supply that the title promises and that survives in these claims only as a precondition for generator detection.

---

AU 2026202180 was published in the Australian Official Journal of Patents on 9 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.