Training a mathematical model for a device using a smart plug

• US 10,586,177 B1
• Filed: 11/02/2018
• Issued: 03/10/2020
• Est. Priority Date: 10/02/2018
• Status: Active Grant

• Current Assignee: Sense Labs, Inc.
• Original Assignee: Sense Labs, Inc.
• Inventors: Choueiter, Ghinwa Fakhri, Petri, Jonah Wyman, Zavaliagkos, George
• Primary Examiner(s): Nilsson, Eric
• Application Number:
  US16/179,619
• Publication Number:
  US 20200104756A1
• Time in Patent Office: 494 Days
• CPC Class Codes: G01R 22/063 related to remote communication
• CPC Class Codes: G01R 22/10 using digital techniques
• CPC Class Codes: G05B 19/042 using digital processors G05B19/05 takes precedence
• CPC Class Codes: G06F 1/28 Supervision thereof, e.g. detecting power-supply failure by out of limits supervision
• CPC Class Codes: G06F 1/30 Means for acting in the event of power-supply failure or interruption, e.g. power-supply fluctuations for resetting only G06F1/24
• CPC Class Codes: G06F 16/2379 Updates performed during online database operations; commit processing
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 3/088 Non-supervised learning, e.g. competitive learning
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 5/046 Forward inferencing; Production systems
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: H02J 13/00007 using the power network as support for the transmission
• CPC Class Codes: H02J 13/00036 the elements or equipment being or involving switches, relays or circuit breakers circuits for indication of single switches H01H9/167
• CPC Class Codes: H04L 67/12 specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
• CPC Class Codes: H04L 67/125 involving control of end-device applications over a network
• CPC Class Codes: H04Q 2209/84 Measuring functions
• CPC Class Codes: H04Q 9/00 : Arrangements in telecontrol or telemetry systems for selectively calling a substation from a main station, in which substation desired apparatus is selected for applying a control signal thereto or for obtaining measured values therefrom
• CPC Class Codes: Y02E 60/00 : Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
• CPC Class Codes: Y04S 10/18 : using switches, relays or circuit breakers, e.g. intelligent electronic devices [IED]
• CPC Class Codes: Y04S 40/121 : using the power network as support for the transmission
• CPC Class Codes: Y04S 40/18 : Network protocols supporting networked applications, e.g. including control of end-device applications over a network

Abstract

A smart plug may provide a smart-plug power monitoring signal that includes information about power consumption of devices connected to the smart plug. The smart-plug power monitoring signal may be used in conjunction with power monitoring signals from the electrical mains of the building for providing information about the operation of devices in the building. For example, the power monitoring signals may be used to (i) determine the main of the house that provides power to the smart plug, (ii) identify devices receiving power from the smart plug, (iii) improve the accuracy of identifying device state changes, and (iv) train mathematical models for identifying devices and device state changes.

Claims

1. A computer-implemented method for training a mathematical model for a first device, the method comprising:

receiving a smart-plug power monitoring signal, wherein the smart-plug power monitoring signal indicates an amount of power provided by a smart plug to one or more devices, the one or more devices comprising the first device;

identifying a plurality of turn-on times corresponding to the smart-plug power monitoring signal transitioning from zero power to non-zero power;

identifying a plurality of turn-off times corresponding to the smart-plug power monitoring signal transitioning from non-zero power to zero power;

obtaining a first-main power monitoring signal using measurements from a sensor that measures an electrical property of a first electrical main of a building;

identifying power events in the first-main power monitoring signal, wherein each power event corresponds to an event time;

clustering the power events into a plurality of clusters;

selecting a first cluster of the plurality of clusters using (i) the plurality of turn-on times and (ii) event times of power events of the first cluster;

selecting a second cluster of the plurality of clusters using (i) the plurality of turn-off times and (ii) event times of power events of the second cluster;

training a first transition model for the first device using one or more of the power events of the first cluster;

training a second transition model for the first device using one or more of the power events of the second cluster; and

identifying state changes of devices using the first transition model and the second transition model.

2. The computer-implemented method of claim 1, wherein selecting the first cluster comprises determining, for each event time of the power events of the first cluster, a closest turn-on time.

3. The computer-implemented method of claim 1, comprising:

creating a state model for the first device comprising two states;

computing a wattage model for the first device;

computing a score indicating a goodness of fit of the wattage model for the first device to the smart-plug power monitoring signal; and

determining to add a third state to the state model for the first device using the score.

4. The computer-implemented method of claim 3, comprising selecting a third cluster of the plurality of clusters using (i) the plurality of turn-on times, (ii) the plurality of turn-off times, and (iii) event times of the power events of the third cluster.

5. The computer-implemented method of claim 4, comprising determining, for each event time of the power events of the third cluster, whether the event time is between a turn-on time and a subsequent turn-off time.

6. The computer-implemented method of claim 3, wherein computing a wattage model for the first device comprises computing an expected amount of power consumed by the first device from the first-main power monitoring signal.

7. The computer-implemented method of claim 6, wherein the expected amount of power is a mean or median of power in the first-main power monitoring signal after the power events of the first cluster.

8. The computer-implemented method of claim 6, wherein the expected amount of power is a mean or median of power in the smart-plug power monitoring signal after the turn-on times.

9. A system for training a mathematical model for a first device, the system comprising:

at least one computer comprising at least one processor and at least one memory, the at least one computer configured to;

receive a smart-plug power monitoring signal, wherein the smart-plug power monitoring signal indicates an amount of power provided by a smart plug to one or more devices, the one or more devices comprising the first device;

identify a plurality of turn-on times corresponding to the smart-plug power monitoring signal transitioning from zero power to non-zero power;

identify a plurality of turn-off times corresponding to the smart-plug power monitoring signal transitioning from non-zero power to zero power;

obtain a first-main power monitoring signal using measurements from a sensor that measures an electrical property of a first electrical main of a building;

identify power events in the first-main power monitoring signal, wherein each power event corresponds to an event time;

cluster the power events into a plurality of clusters;

select a first cluster of the plurality of clusters using (i) the plurality of turn-on times and (ii) event times of power events of the first cluster;

select a second cluster of the plurality of clusters using (i) the plurality of turn-off times and (ii) event times of power events of the second cluster;

train a first transition model for the first device using one or more of the power events of the first cluster;

train a second transition model for the first device using one or more of the power events of the second cluster; and

identify state changes of devices using the first transition model and the second transition model.

10. The system of claim 9, wherein the at least one computer is configured to deploy the first transition model and the second transition model to a second building.

11. The system of claim 9, wherein the at least one computer is configured to select the first cluster by determining, for each event time of the power events of the first cluster, a closest turn-on time.

12. The system of claim 9, wherein the at least one computer is configured to:

create a state model for the first device comprising two states;

compute a wattage model for the first device;

compute a score indicating a goodness of fit of the wattage model for the first device to the smart-plug power monitoring signal; and

determine to add a third state to the state model for the first device using the score.

13. The system of claim 12, wherein the at least one computer is configured to determine to add a fourth state to the state model for the first device.

14. The system of claim 12, wherein the at least one computer is configured to compute the score indicating the goodness of fit of the wattage model by computing a distance between the wattage model and a plurality of portions of the smart-plug power monitoring signal.

15. The system of claim 9, wherein the at least one computer is configured to:

compute features for each power event of the power events; and

wherein the at least one computer is configured to cluster the power events using the features.

16. The system of claim 15, wherein the features computed for a power event comprise one or more of a match between the power event and a template, Fourier coefficients, neural network features, or a change in power, current, voltage, or phase before and after the power event.

17. One or more non-transitory, computer-readable media comprising computer executable instructions that, when executed, cause at least one processor to perform actions comprising:

receiving a smart-plug power monitoring signal, wherein the smart-plug power monitoring signal indicates an amount of power provided by a smart plug to one or more devices, the one or more devices comprising a first device;

identifying a plurality of turn-on times corresponding to the smart-plug power monitoring signal transitioning from zero power to non-zero power;

identifying a plurality of turn-off times corresponding to the smart-plug power monitoring signal transitioning from non-zero power to zero power;

obtaining a first-main power monitoring signal using measurements from a sensor that measures an electrical property of a first electrical main of a building;

identifying power events in the first-main power monitoring signal, wherein each power event corresponds to an event time;

clustering the power events into a plurality of clusters;

selecting a first cluster of the plurality of clusters using (i) the plurality of turn-on times and (ii) event times of power events of the first cluster;

selecting a second cluster of the plurality of clusters using (i) the plurality of turn-off times and (ii) event times of power events of the second cluster;

training a first transition model for the first device using one or more of the power events of the first cluster;

training a second transition model for the first device using one or more of the power events of the second cluster; and

identifying state changes of devices using the first transition model and the second transition model.

18. The computer-readable media of claim 17, wherein selecting the first cluster comprises determining, for each event time of the power events of the first cluster, a closest turn-on time.

19. The computer-readable media of claim 17, wherein clustering the power events comprises using hierarchical clustering, centroid-based clustering, or density-based clustering.

20. The computer-readable media of claim 17, wherein the actions comprise:

computing features for each power event of the power events; and

wherein clustering the power events comprises clustering the power events using the features.