Solution of multistage decision problems considering subjectively evaluated risks of changes in problems

2014 Fiscal Year Final Research Report

• Project/Area Number: 24560499
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: System engineering
• Research Institution: Kyushu University
• Principal Investigator: MURATA Junichi 九州大学, システム情報科学研究科(研究院, 教授 (60190914)
• Co-Investigator(Renkei-kenkyūsha): TAKANO Hirotaka 九州大学, 大学院システム情報科学研究院, 助教 (50435426)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 最適化 / 不確定性 / リスク / 機械学習 / 再生可能エネルギー

Outline of Final Research Achievements

Methods have been developed for multistage decision problems with changes.
A multistage decision problem finds a sequence of appropriate alternatives such that their overall evaluation is optimized. The typical evaluation is finding a shortest path in a network of streets. The changes considered are those in the set of alternatives which corresponds to street closures in street networks.
The following have been developed: a method for subjective evaluation of risks based on the probability of and the damages caused by changes, a reinforcement learning algorithm that solves the problem considering the subjective risks, and learning classifier systems that acquire multiple solutions and switch one from another in response to the changes. Also, operation planning schema have been developed for power systems with photovoltaic generation units whose outputs changes non-deterministically.

Free Research Field

システム工学