Systems and methods for maximizing expected utility of signal injection test patterns in utility grids

• US 9,922,293 B2
• Filed: 07/14/2015
• Issued: 03/20/2018
• Est. Priority Date: 07/17/2014
• Status: Active Grant

• Current Assignee: 3M Innovative Properties Company (3M Company)
• Original Assignee: 3M Innovative Properties Company (3M Company)
• Inventors: Brooks, Brian E., Lu, Yang, Tio, Andrew T., Ong, Chong Yang, Benoit, Gilles J.
• Primary Examiner(s): VINCENT, DAVID ROBERT
• Application Number:
  US15/324,809
• Publication Number:
  US 20170206467A1
• Time in Patent Office: 980 Days
• Field of Search: 706 12, 706 45
• CPC Class Codes: G05B 23/0256 injecting test signals and analyzing monitored process response, e.g. injecting the test signal while interrupting the normal operation of the monitored system; superimposing the test signal onto a control signal during normal operation of the monitored system
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 50/06 Energy or water supply

Abstract

Methods and systems for implementing experimental trials on utility grids. Variations in grid parameters are selected to introduce into utility grids to improve the value of learning from each experimental trial and promoting improved utility grid performance by computing expected values for both learning and grid performance. Those trials are used to manage the opportunity costs and constraints that affect the introduction of variations into utility grid parameters and the generation of valid data that can be attributed to particular variations in utility grid parameters.

Claims

1. A method for injecting signals into a utility grid, comprising:

receiving a spatial reach and a temporal reach for each signal injection of a plurality of signal injections;

computing a learning value for each signal injection in the plurality of signal injections;

computing an expected effect value for each signal injection in the plurality of signal injections;

selecting, based on the learning values and expected effect values, a set of signal injections wherein the spatial reach and temporal reach of each signal injection do not both overlap the spatial reach and temporal reach of another signal injection in the set; and

injecting the selected set of signal injections into a utility grid,

wherein the signal injections are changes in the state of grid controls,

wherein the grid control is a capacitor bank.

2. The method of claim 1, further comprising measuring the utility grid response to the injected signals.

3. The method of claim 2, further comprising associating data from the sensors with signal injections based on the time and location of the sensor data and the spatial reach and temporal reach of the signal injections.

4. The method of claim 1, wherein the learning value is computed based on a number of belief states that may be falsified by grid response to the signal injection.

5. The method of claim 1, wherein the learning value is computed based on a predicted change in the width of confidence intervals for grid response to the signal injection.

6. The method of claim 1, wherein the expected effect value is computed based on a database of the effects of prior signal injections.

7. The method of claim 1, wherein multiple signal injections are concurrently implemented on the utility grid.

8. The method of claim 1, wherein the signal injections are coordinated by a Partially Observable Markov Decision Process.

9. A method for injecting signals into a utility grid, comprising:

receiving a spatial reach and a temporal reach for each signal injection of a plurality of signal injections;

computing a learning value for each signal injection in the plurality of signal injections;

computing an expected effect value for each signal injection in the plurality of signal injections;

selecting, based on the learning values and expected effect values, a set of signal injections wherein the spatial reach and temporal reach of each signal injection do not both overlap the spatial reach and temporal reach of another signal injection in the set and injecting the selected set of signal injections into a utility grid,

wherein the signal injections are dispatching of grid personnel to perform a task.

10. A utility grid system, comprising:

a spatial reach memory, configured to store the spatial reach for each of a plurality of signal injections;

a temporal reach memory, configured to store the temporal reach for each of a plurality of signal injections;

an expected effect value processor, configured to compute an expected effect value for a signal injection;

a learning value processor, configured to compute a learning value for a signal injection;

a coordination processor, configured to generate a set of signal injections where the spatial and temporal reaches do not both overlap for any signal injections in the set;

a plurality of utility grid controls; and

sensors located along the utility grid,

wherein the sensors are electrical sensors.

11. The utility grid system of claim 10, further comprising a processor configured to associate sensor data with signal injections.

12. The utility grid system of claim 10, wherein the expected effect value processor computes expected effect value by predicting the impact of a signal injection on the utility grid.

13. The utility grid system of claim 10, wherein the learning value processor computes learning value by determining the number of belief states that may be falsified by a signal injection.

14. The utility grid system of claim 10, wherein the coordination processor coordinates the signal injections using a graphical model.

15. The utility grid system of claim 10, wherein the grid controls concurrently implement multiple signal injections.

16. A utility grid system, comprising:

a spatial reach memory, configured to store the spatial reach for each of a plurality of signal injections;

a temporal reach memory, configured to store the temporal reach for each of a plurality of signal injections;

an expected effect value processor, configured to compute an expected effect value for a signal injection;

a learning value processor, configured to compute a learning value for a signal injection;

a coordination processor, configured to generate a set of signal injections where the spatial and temporal reaches do not both overlap for any signal injections in the set; and

a plurality of utility grid controls; and

sensors located along the utility grid,

wherein the sensors are methane sensors.