Intelligent Wind Farm Control via Grouping-Based Reinforcement Learning

Abstract

How to cite:Please refer to published version for the most recent bibliographic citation information. If a published version is known of, the repository item page linked to above, will contain details on accessing it.

Reinforcement Learning-Based Wind Farm Control: Toward Large Farm Applications via Automatic Grouping and Transfer Learning

“…It is a quantitative way to show the influence of a single turbine's power generating process on the flow field around it. For any spatial position H in the farm, we use IF W Ti→H to denote the influence factor of turbine WT i with respect to the position H. Following [30], the specific definition of IF W Ti→H is…”

“…systems with unknown models). Applying RL in wind farm operations has become a cutting-edge research area, and its feasibility has been proved in recent studies [23], [24], [25], [26], [27], [28], [29], [30], [31]. For example, a model-free approach for wind farm power optimization was introduced in [23] via the deep Q-network algorithm [32].…”

“…It is noteworthy that this paper is partially built upon the conference version in [30], while significant new contributions have been made from the following several aspects: (1) Based on the adjacency matrix and a restricted depth-first search approach, a solid method (Algorithms 1 & 2) is proposed in this paper to generate turbine groups automatically -to our best knowledge, this is for the first time an automatic grouping strategy is proposed for wind farm operational tasks. In contrast, Ref.…”

“…In contrast, Ref. [30] requires manual grouping. (2) Significantly different from [30], a new RL-based wind farm control method is developed in this paper via multi-agent proximal policy optimization.…”

“…[30] requires manual grouping. (2) Significantly different from [30], a new RL-based wind farm control method is developed in this paper via multi-agent proximal policy optimization. Compared with [30], this new method is more suitable for handling wind farm control problems, especially in cooperation with the automatic grouping algorithm.…”