Machine learning in fusion reactors

Patent metadata

Field	Value
Publication number	US-11587690-B2
Application number	US-201816627599-A
Country	US
Kind code	B2
Filing date	Jul 4, 2018
Priority date	Jul 6, 2017
Publication date	Feb 21, 2023
Grant date	Feb 21, 2023

Abstract

A method of controlling a plasma in a nuclear fusion reactor. The nuclear fusion reactor comprises sensors and plasma control inputs. An initial control model is provided, relating readings of at least a subset of the sensors to control of the plasma control inputs. A control loop is performed, comprising: operating the plasma control inputs in dependence upon the sensors according to the control model; determining correlations between readings of each of the sensors, and/or between readings of the sensors and states of the plasma control inputs; and adjusting the control model based on the determined correlations.

First claim

The invention claimed is:

1. A method of controlling a plasma within a plasma vessel of a nuclear fusion reactor, the nuclear fusion reactor including one or more plasma control inputs, one or more primary sensors and one or more secondary sensors, and wherein the plasma control inputs are configured to adjust the properties of the plasma, the method comprising:
providing an initial control model relating readings of only the one or more primary sensors to control of the plasma control inputs;
performing a control loop including:
controlling the plasma using the one or more plasma control inputs in dependence upon the one or more primary sensors and/or one or more secondary sensors according to the control model;
determining, using a machine learning algorithm, correlations between readings of the one or more secondary sensors and readings of the one or more primary sensors, and/or between readings of the one or more secondary sensors and states of the one or more plasma control inputs; and
adjusting the control model based on the correlations determined by the machine learning algorithm, wherein adjusting the control model includes adding a secondary sensor to the control model when a correlation has been found between readings of the one or more secondary sensors and readings of one or more primary sensors, and/or between readings of the one or more secondary sensors and states of the one or more plasma control inputs, and
controlling the plasma according to the adjusted control model;
wherein the one or more plasma control inputs control external heating and current drive, position control coils, fueling systems, and disruption control systems, the one or more primary sensors monitor the plasma and the one or more secondary sensors monitor aspects of the reactor other than the plasma.

2. The method according to claim 1 , wherein adjusting the control model includes one or more of:
adjusting weightings of one or more primary sensors and/or one or more secondary sensors within the control model on the basis of the determined correlations, and
removing one or more primary sensors and/or one or more secondary sensors from the control model if it is determined that there is no longer sufficient correlation between that sensor and other primary sensors or secondary sensors and/or the plasma control inputs.

3. The method according to claim 1 , wherein the primary sensors comprise any one or more of:
magnetic diagnostics;
spectroscopic instruments;
optical sensors;
bolometric systems;
microwave diagnostics.

4. The method according to claim 1 , wherein the secondary sensors comprise any one or more of:
stress sensors;
strain sensors;
microphones;
vibration sensors;
temperature sensors; and
neutron detectors.

5. The method according to claim 1 , wherein the plasma control inputs comprise any one or more of:
magnetic field coils;
poloidal field coils;
toroidal field coils;
a heating system;
a current drive system; and
a fuel input.

6. The method according to claim 1 , wherein the nuclear fusion reactor is a spherical tokamak.

7. A controller for a nuclear fusion reactor, the controller being configured to interface with sensors and plasma control inputs of the nuclear fusion reactor, and to perform a method according to claim 1 .

8. A non-transitory computer-readable medium storing instructions that, when executed by a controller for a nuclear fusion reactor, cause the controller to perform a method according to claim 1 .

9. The method according to claim 1 , wherein the one or more secondary sensors are configured to monitor at least one selected from the group consisting of structural stresses, structural strains, temperature of a superconducting magnet, noise, and neutron flux.

10. The method according to claim 1 , further comprising initialising the reactor in a low neutron production mode, and switching to a power generation mode following a training period.

11. The method according to claim 10 , wherein the training period finishes:
after a predetermined time has elapsed; or
when a given accuracy and/or confidence level of the control model has been achieved.

12. A method of controlling a position, temperature and/or current of a plasma within a plasma vessel of a nuclear fusion reactor, the nuclear fusion reactor including sensors and plasma control inputs, wherein the sensors include one or more primary sensors and one or more secondary sensors, and wherein the plasma control inputs are configured to adjust the properties of the plasma, the method comprising:
providing an initial control model relating readings of only the one or more primary sensors to control of the plasma control inputs; and
performing a control loop including:
controlling the position, temperature, current, density and/or shape of the plasma using the plasma control inputs in dependence upon the one or more primary and/or one or more secondary sensors according to the control model,
determining, using a machine learning algorithm, correlations between readings of the one or more secondary sensors and readings of the one or more primary sensors, and/or between readings of the one or more secondary sensors and states of the plasma control inputs,
adjusting the control model based on the correlations determined by the machine learning algorithm, wherein adjusting the control model includes adding a secondary sensor to the control model when a correlation has been found between readings of the one or more secondary sensor and readings of the one or more primary sensors and/or between readings of the one or more secondary sensors and states of the plasma control inputs, and
controlling the position, temperature and/or current of the plasma based on the adjusted control model;
wherein the one or more plasma control inputs control external heating and current drive, position control coils, fueling systems, and disruption control systems, the one or more primary sensors monitor the plasma and the one or more secondary sensors monitor aspects of the reactor other than the plasma.

Assignees

• Tokamak Energy Ltd

Inventors

• Mcnamara Steven Antony Milton
• Buxton Peter
• Kingham David

Classifications

• Y02E30/10

  Nuclear fusion reactors · CPC title

• G21B1/00 Primary

  Thermonuclear fusion reactors · CPC title

• G21D3/001 Primary

  Computer implemented control · CPC title

• G21D3/08

  Regulation of any parameters in the plant · CPC title

• G21B1/057 Primary

  Tokamaks · CPC title