Reinforcement learning-enhanced genetic algorithm for wind farm layout optimization

TL;DR

This work tackles wind farm layout optimization by addressing GA parameter sensitivity with a reinforcement learning–enhanced GA (RLGA) that dynamically selects genetic operators via Q-learning. The framework uses the Jensen wake model with wake superposition to compute turbine deficits and a cost-based objective $f_{obj}=P_{total}/cost$, with fitness $F=1/(f_{obj}-f_{obj,ideal})$; operator choices are guided by a Q-table updated through rewards $R_t=F_t-F_{t-1}$ and Bellman updates $Q_{t+1}(S_t,A_t)=Q_t(S_t,A_t)+\alpha(R_t+\gamma\max_a Q(S_{t+1},a)-Q_t(S_t,A_t))$. RLGA dynamically selects two mating options, four crossover types, and four mutation rates, enabling faster convergence and better performance on complex WFLO problems, including large farms with diverse layouts and wind conditions. The results show that RLGA matches GA for simple, aligned and staggered layouts while outperforming GA for sunflower and unstructured layouts, and achieves up to approximately three times the convergence efficiency for large-scale problems, highlighting its practical potential for scalable WFLO and future multi-objective extensions.

Abstract

A reinforcement learning-enhanced genetic algorithm (RLGA) is proposed for wind farm layout optimization (WFLO) problems. While genetic algorithms (GAs) are among the most effective and accessible methods for WFLO, their performance and convergence are highly sensitive to parameter selections. To address the issue, reinforcement learning (RL) is introduced to dynamically select optimal parameters throughout the GA process. To illustrate the accuracy and efficiency of the proposed RLGA, we evaluate the WFLO problem for four layouts (aligned, staggered, sunflower, and unstructured) under unidirectional uniform wind, comparing the results with those from the GA. RLGA achieves similar results to GA for aligned and staggered layouts and outperforms GA for sunflower and unstructured layouts, demonstrating its efficiency. The sunflower and unstructured layouts' complexity highlights RLGA's robustness and efficiency in tackling complex problems. To further validate its capabilities, we investigate larger wind farms with varying turbine placements ($Δx = Δy = 5D$ and 2$D$, where $D$ is the wind turbine diameter) under three wind conditions: unidirectional, omnidirectional, and non-uniform, presenting greater challenges. The proposed RLGA is about three times more efficient than GA, especially for complex problems. This improvement stems from RL's ability to adjust parameters, avoiding local optima and accelerating convergence.

Figures (12)

• Figure 1: The schematic (left) and contour (right) of the Jensen single wake model.
• Figure 2: Overlapped area between a wind turbine rotor and a wake stream.
• Figure 3: Block diagram schematic for RL.
• Figure 4: The flow chart of the RLGA method with GA and RL algorithms shown in pink and blue boxes, respectively.
• Figure 5: The wind rose for cases the omnidirectional uniform wind (left) and spread non-uniform wind (right).