Title:
REINFORCEMENT LEARNING METHOD, REINFORCEMENT LEARNING PROGRAM, AND REINFORCEMENT LEARNING APPARATUS
Document Type and Number:
Japanese Patent JP2020119139
Kind Code:
A
Abstract:
To improve learning efficiency by reinforcement learning.SOLUTION: A value function learning section 403 executes a unit learning step and learns a value function on the basis of a received state of a wind power generation facility 400, a reward of the wind power generation facility 400, and action to the wind power generation facility 400. An experience degree calculation section 404 updates an experience degree function on the basis of the received state of the wind power generation facility 400, the reward of the wind power generation facility 400, and the action to the wind power generation facility 400. The experience degree calculation section 404 calculates an experience degree of a state or action at this time and an experience degree of a state or action at the other time for the wind power generation facility 400 on the basis of the experience degree function. A value function correction section 405 determines whether or not to update the value function more on the basis of the value function and the experience degree. The value function correction section 405 updates the value function using monotonicity on the basis of the value function and the experience degree when determining to update the value function.SELECTED DRAWING: Figure 4
More Like This:
Inventors:
SHIGEZUMI JUNICHI
IWANE HIDENAO
YANAMI HITOSHI
IWANE HIDENAO
YANAMI HITOSHI
Application Number:
JP2019008512A
Publication Date:
August 06, 2020
Filing Date:
January 22, 2019
Export Citation:
Assignee:
FUJITSU LTD
International Classes:
G06N20/00
Domestic Patent References:
| JP2018005739A | 2018-01-11 |
Other References:
PARR, R ET AL.: ""An analysis of linear models, linear value-function approximation, and feature selection for reinfo", ICML '08: PROCEEDINGS OF THE 25TH INTERNATIONAL CONFERENCE ON MACHINE LEARNING [ONLINE], JPN7022004312, 2008, US, pages 752 - 759, XP058106385, ISSN: 0004872400, DOI: 10.1145/1390156.1390251
WEI, C ET AL.: ""An Adaptive Network-Based Reinforcement Learning Method for MPPT Control of PMSG Wind Energy Conver", IEEE TRANSACTIONS ON POWER ELECTRONICS [ONLINE], vol. 31, no. 11, JPN6022037814, 2016, pages 7837 - 7848, XP011615255, ISSN: 0004872401, DOI: 10.1109/TPEL.2016.2514370
WEI, C ET AL.: ""An Adaptive Network-Based Reinforcement Learning Method for MPPT Control of PMSG Wind Energy Conver", IEEE TRANSACTIONS ON POWER ELECTRONICS [ONLINE], vol. 31, no. 11, JPN6022037814, 2016, pages 7837 - 7848, XP011615255, ISSN: 0004872401, DOI: 10.1109/TPEL.2016.2514370
Attorney, Agent or Firm:
Akinori Sakai