US 11,809,164 B2
Integrating machine learning into control systems for industrial facilities
Jim Gao, Menlo Park, CA (US); Christopher Gamble, London (GB); Amanda Gasparik, Philadelphia, PA (US); Vedavyas Panneershelvam, London (GB); David Barker, Reading (GB); Dustin Reishus, Mountain View, CA (US); Abigail Ward, Mountain View, CA (US); Jerry Luo, Mountain View, CA (US); Brian Kim, Sunnyvale, CA (US); Mark Schwabacher, Santa Clara, CA (US); Stephen Webster, San Mateo, CA (US); Timothy Jason Kieper, Beacon, NY (US); Daniel Fuenffinger, Bellevue, NE (US); and Zakerey Bennett, London (GB)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Feb. 25, 2022, as Appl. No. 17/681,652.
Application 17/681,652 is a continuation of application No. 16/654,978, filed on Oct. 16, 2019, abandoned.
Application 16/654,978 is a continuation of application No. PCT/US2018/029611, filed on Apr. 26, 2018.
Claims priority of provisional application 62/490,544, filed on Apr. 26, 2017.
Prior Publication US 2022/0179401 A1, Jun. 9, 2022
Int. Cl. H02J 3/46 (2006.01); G06Q 10/04 (2023.01); G06Q 10/10 (2023.01); G05B 19/4155 (2006.01); G06N 20/00 (2019.01)
CPC G05B 19/4155 (2013.01) [G06N 20/00 (2019.01); G05B 2219/40499 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, from a machine learning system and by a control system for a controllable facility, a first optimal controllable facility setting slate generated by a machine learning system for the controllable facility in a first state at a first time, wherein the first optimal controllable facility setting slate includes first values for setting a plurality of controllable facility controls;
determining, by the control system for the controllable facility, that the first values in the first optimal controllable facility setting slate, once adopted by the controllable facility in the first state, will not result in unstable conditions in the controllable facility;
in response to determining that the first values will not result in unstable conditions in the controllable facility, adopting the first values in the first optimal controllable facility setting slate for controlling the controllable facility;
receiving, from the machine learning system and by the control system for the controllable facility, a second optimal controllable facility setting slate generated by the machine learning system for the controllable facility in a second state at a second time, wherein the second optimal controllable facility setting slate includes second values for setting the plurality of controllable facility controls;
determining, by the control system for the controllable facility, that the second values in the second optimal controllable facility setting slate, once adopted by the controllable facility in the second state, will result in unstable conditions in the controllable facility; and
in response to determining that the second values will result in unstable conditions in the controllable facility, adopting settings provided by a default control system for controlling the controllable facility.