Systems and methods for predicting outages in electrical networks based on voltage measurements from SCADA-enabled devices

1 . A method, comprising:

receiving sensor data measured by a plurality of SCADA-enabled sensors associated with an electrical grid;

identifying a plurality of momentary outages (1) associated with a component of the electrical grid, (2) having a duration that is less than a predefined duration, and (3) represented by the sensor data;

predicting, for each momentary outage from the plurality of momentary outages, a time period until that momentary outage transforms into a sustained outage that has a duration that is greater than the predefined duration, to produce a plurality of time periods;

identifying a plurality of possible sustained outages associated with the electrical grid based on the plurality of time periods being less than a predefined time period; and

predicting a sustained outage likelihood from a plurality of sustained outage likelihoods for each possible sustained outage from the plurality of possible sustained outages based on the plurality of momentary outages.

2 . The method of claim 1 , wherein the component includes an interrupting device.

3 . The method of claim 1 , wherein predicting the plurality of sustained outage likelihoods includes:

associating the plurality of momentary outages with a plurality of probable causes; and

predicting, for each possible sustained outage from the plurality of possible sustained outages, a probable cause from the plurality of probable causes, the plurality of sustained outage likelihoods being calculated based on the plurality of probable causes.

4 . The method of claim 1 , wherein:

each time period from the plurality of time periods represents a number of days until that momentary outage transforms into a possible sustained outage from the plurality of possible sustained outages.

5 . The method of claim 1 , wherein the sensor data includes data from Advanced Metering Infrastructure (AMI).

6 . The method of claim 1 , wherein the sensor data includes data from Advanced Metering Infrastructure (AMI) sensors and data from Automatic Meter Reader (AMR) sensors.

7 . The method of claim 1 , further comprising:

sending a signal to cause a display of a user interface that shows an interactive visualization of at least some momentary outages from the plurality of momentary outages and at least one possible sustained outage from the plurality of possible sustained outages, the interactive visualization including a representation of at least one of circuits of the electrical grid, SCADA-enabled devices, changes over time within the electrical grid, customers of the electrical grid, or geography of the electrical grid.

8 . The method of claim 1 , wherein:

the identifying the plurality of possible sustained outages includes providing momentary outage data as input to a first machine learning model to identify the plurality of possible sustained outages, the momentary outage data representing the plurality of momentary outages; and

the predicting the sustained outage likelihood for each possible sustained outage from the plurality of possible sustained outages includes providing the momentary outage data as input to a second machine learning model to predict, for each possible sustained outage from the plurality of possible sustained outages, the sustained outage likelihood (1) for that possible sustained outage and (2) indicating a confidence that a momentary outage from the plurality of momentary outages will transition to that possible sustained outage.

9 . The method of claim 1 , wherein:

a first possible sustained outage from the plurality of possible sustained outages is associated with a first phase from a plurality of phases of the electrical grid; and

a second possible sustained outage from the plurality of possible sustained outages is associated with a second phase, different from the first phase, from the plurality of phases of the electrical grid.

10 . The method of claim 1 , further comprising:

receiving geographic location data associated with the plurality of SCADA-enabled sensors;

identifying a weather station based on the geographic location data; and

in response to the identifying the weather station, receiving weather data that is recorded at the weather station, the plurality of possible sustained outages being predicted based further on the weather data.

11 . The method of claim 1 , further comprising:

receiving an image that depicts infrastructure that is associated with the electrical grid; and

providing the image as input to a machine learning model to detect an anomaly associated with the infrastructure,

the identifying the plurality of possible sustained outages being based further on the anomaly.

12 . A non-transitory, processor-readable medium storing instructions that, when executed by a processor, cause the processor to:

receive sensor data measured by a plurality of SCADA-enabled sensors associated with an electrical grid;

identify a plurality of momentary outages (1) associated with a component of the electrical grid, (2) having a duration that is less than a predefined duration, and (3) represented by the sensor data;

predict, for each momentary outage from the plurality of momentary outages, a time period until that momentary outage transforms into a sustained outage that has a duration that is greater than the predefined duration, to produce a plurality of time periods;

identify a plurality of possible sustained outages associated with the electrical grid based on the plurality of time periods being less than a predefined time period; and

predict a sustained outage likelihood from a plurality of sustained outage likelihoods for each possible sustained outage from the plurality of possible sustained outages based on the plurality of momentary outages.

13 . The non-transitory, processor-readable medium of claim 12 , wherein the component includes an interrupting device.

14 . The non-transitory, processor-readable medium of claim 12 , wherein the instructions to cause the processor to predict the plurality of sustained outage likelihoods include instructions to cause the processor to:

associate each momentary outage from the plurality of momentary outages with a probable cause from a plurality of probable causes; and

predict, for each possible sustained outage from a plurality of possible sustained outages, a probable cause from the plurality of probable causes, the plurality of sustained outage likelihoods being calculated based on the plurality of probable causes.

15 . The non-transitory, processor-readable medium of claim 12 , wherein each time period from the plurality of time periods represents a number of days until that momentary outage transforms into a possible sustained outage from the plurality of possible sustained outages, the non-transitory, processor-readable medium further storing instructions to cause the processor to:

calculate, for each momentary outage from the plurality of momentary outages, a confidence score for that momentary outage and that is proportional to the number of days for that momentary outage, to produce a plurality of confidence scores, the plurality of possible sustained outages being predicted based on the plurality of confidence scores.

16 . The non-transitory, processor-readable medium of claim 12 , wherein the sensor data includes data from Advanced Metering Infrastructure (AMI).

17 . The non-transitory, processor-readable medium of claim 12 , wherein the sensor data includes data from Advanced Metering Infrastructure (AMI) sensors and data from Automatic Meter Reader (AMR) sensors.

18 . The non-transitory, processor-readable medium of claim 12 , further storing instructions to cause the processor to:

send a signal to cause a display of a user interface that shows an interactive visualization of at least some momentary outages from the plurality of momentary outages and at least one possible sustained outage from the plurality of possible sustained outages, the interactive visualization including a representation of at least one of circuits of the electrical grid, SCADA-enabled devices, changes over time within the electrical grid, customers of the electrical grid, or geography of the electrical grid.

19 . The non-transitory, processor-readable medium of claim 12 , wherein:

the instructions to cause the processor to identify the plurality of possible sustained outages include instructions to cause the processor to provide momentary outage data as input to a first machine learning model to identify the plurality of possible sustained outages, the momentary outage data representing the plurality of momentary outages; and

the instructions to cause the processor to predict the sustained outage likelihood for each possible sustained outage from the plurality of possible sustained outages include instructions to cause the processor to provide the momentary outage data as input to a second machine learning model to predict, for each possible sustained outage from the plurality of possible sustained outages, the sustained outage likelihood (1) for that possible sustained outage and (2) indicating a confidence that a momentary outage from the plurality of momentary outages will transition to that possible sustained outage.

20 . The non-transitory, processor-readable medium of claim 12 , further storing instructions to cause the processor to:

receive an image that depicts infrastructure that is associated with the electrical grid; and

provide the image as input to a second machine learning model to detect an anomaly associated with the infrastructure, the plurality of possible sustained outages being predicted based further on the anomaly.

21 . The method of claim 1 , wherein:

the identifying the plurality of possible sustained outages includes providing an indication of the plurality of momentary outages as input to a machine learning model to identify the plurality of possible sustained outages; and

the method further comprises:

receiving field reported outage data; and

separating the field reported outage data based on a duration threshold to produce field reported momentary outage data and field reported sustained outage data, the machine learning model being trained based on the field reported momentary outage data and the field reported sustained outage data.

22 . The method of claim 1 , further comprising:

providing an indication of the plurality of momentary outages as input to a machine learning model to predict a time duration between (1) at least one momentary outage from the plurality of momentary outages and (2) at least one possible sustained outage from the plurality of possible sustained outages.