強化学習による洋上風力タービン群の微風時発電効率改善に関する研究

概要

論文の詳細を見る
Recently wind energy is getting a lot of attention as one of renewable energy sources. This is because electric-power-generation efficiency has been improved reasonably over the last decade based on modern control technologies. In particular, several methods of power assistant for a single wind turbine are proposed for the electric power generation in slight wind condition. However, there is very little discussion on power assistant in the case of several wind turbines. Also it is desirable to set them not in land but offshore because of space and noise problems. Therefore, the paper considers a power assistant problem in an electric power network system of offshore wind turbines in slight wind condition, where one turbine can exchange some electric power with other turbines in order to realize the most efficient generation of total electric power in a cooperative manner. A methodology of the power assistant is discussed from the viewpoint of AI technology, that is, "Reinforcement Learning". Using computer simulations based on the proposed method, it is shown that electric-power-generation efficiency can be fairly improved compared with the case of uncooperative wind turbines.
社団法人日本船舶海洋工学会の論文
2004-08-31