Graph-based Deep Reinforcement Learning for Wind Farm Set-Point Optimisation

Helen M Sheehan*, Daniel J Poole, Telmo de Menezes e Silva Filho, Ervin A Bossanyi, Lars Landberg

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference Contribution (Conference Proceeding)

Abstract

Wake steering is a form of wind farm control in which upstream turbines are deliberately yawed to misalign with the free-stream wind in order to prevent their wakes from impacting turbines further downstream. This technique can give a net increase in power generated by an array of turbines compared to greedy control, but the optimisation of multiple turbine set-points under varying wind conditions can be infeasibly complex for traditional, whitebox models. In this work, a novel deep reinforcement learning method combining the standard Deep Deterministic Policy Gradient algorithm with a graph representation of potential inter-turbine wake connections was trained to apply wake steering to an array of nine turbines under varying wind directions. The method demonstrated strong performance for wind directions with large potential farm power gains. A steady-state wind farm solver was used, employing a “quasi-dynamic” approach to sampling wind directions, to achieve an additional 47 MW (6.5%) power over four wind directions compared to greedy control.
Original languageEnglish
Title of host publicationThe Science of Making Torque from Wind (TORQUE 2024)
PublisherInstitute of Physics
Publication statusAccepted/In press - 19 Apr 2024
EventThe Science of Making Torque from Wind - Firenze Fiera - Palazzo degli Affari, Florence, Italy
Duration: 29 May 202431 May 2024
https://www.torque2024.eu/

Publication series

NameJournal of Physics: Conference Series
PublisherIOP
ISSN (Print)1742-6596

Conference

ConferenceThe Science of Making Torque from Wind
Abbreviated titleTORQUE 2024
Country/TerritoryItaly
CityFlorence
Period29/05/2431/05/24
Internet address

Fingerprint

Dive into the research topics of 'Graph-based Deep Reinforcement Learning for Wind Farm Set-Point Optimisation'. Together they form a unique fingerprint.

Cite this