Nikolay DImitrov

Teodor Åstrand, Marcus Binder Nilsen, Iasonas Tsaklis et al.

This study presents a multi-agent reinforcement learning (MARL) framework for load-constrained wind farm flow control (WFFC). While wake steering can enhance total wind farm power, it often introduces increased structural loads on downstream turbines. To address this, we integrate an Independent Soft Actor-Critic (I-SAC) architecture with a data-driven, local inflow sector-averaged surrogate model to provide real-time estimates of Damage Equivalent Loads (DELs). By incorporating these estimates into a shaped reward function, turbine-specific agents are trained to maximize power production while adhering to specific load-increase thresholds ($Δ_{max}$) of 10%, 20%, and 30% relative to a baseline controller. The framework is implemented within the WindGym environment using the DYNAMIKS flow solver with Dynamic Wake Meandering (DWM) model to capture non-stationary wake physics. Results indicate that the MARL agents successfully learn collaborative policies that prioritise power gain while actively retreating from high-DEL control strategies.

Marcus Binder Nilsen, Julian Quick, Tuhfe Göçmen et al.

Reinforcement learning (RL) offers a promising approach for adaptive wind farm flow control, yet its practical deployment is hindered by slow training convergence and poor initial performance, factors that could translate to years of reduced power output if an untrained agent were deployed directly. This work investigates whether domain knowledge from steady-state wake models can accelerate RL training and improve initial controller performance. We propose a pretraining methodology in which expert demonstrations are generated by deploying a PyWake-based steady-state optimizer within a dynamic wake simulation (WindGym), then used to initialize both the actor and critic networks of a Soft Actor-Critic agent via behavior cloning. Experiments on a 2x2 wind farm show that pretraining eliminates the costly initial learning phase: while an untrained agent underperforms the greedy zero-yaw baseline by approximately 12%, pretraining raises initial performance to near-baseline levels. During online fine-tuning, all configurations converge within 250,000 environment steps to achieve similar performance, ultimately exceeding that of a lookup-table controller, which reaches approximately 7% power gain after 500,000 steps.

Audrey Bürki, Emiel van den Hoven, Niels O. Schiller et al.

It has been proposed that the order in which words are prepared for production depends on the speaker's language. When producing the translation equivalent of the small cat, speakers of German or Dutch select the gender-marked determiner at a relatively early stage of production. Speakers of French or Italian postpone the encoding of a determiner or adjective until the phonological form of the noun is available. Hence, even though the words are produced in the same order (e.g., die kleine Katze in German, le petit chat in French), they are not planned in the same order and might require different amounts of advanced planning prior to production onset. This distinction between early and late selection languages was proposed to account for the observation that speakers of Germanic and Slavic languages, but not of Romance languages, are slower to name pictures in the context of a distractor word of a different gender. Meta-analyses are conducted to provide the first direct test of this cross-linguistic difference and to test a prediction of the late selection hypothesis. They confirm the existence of the gender congruency effect in German/Slavic languages and its absence in Romance languages when target and distractor words are presented simultaneously. They do not allow confirming the hypothesis that in the latter languages, a similar effect emerges when the presentation of the distractor is delayed. Overall, these analyses confirm the cross-linguistic difference but show that the evidence available to date is not sufficient to confirm or reject the late selection hypothesis as an explanation of this difference. We highlight specific directions for future research.