Wind Farm Turbine Coordination Using Multi-Agent RL
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wind farm control algorithms are ineffective in managing large wind farms with varying wind conditions, as they assume constant wind directions and optimize turbines individually without considering interactions between turbines, leading to reduced energy production and increased operational expenditures.
Innovation Solution
A method using multi-agent reinforcement learning to train a control model that determines actions for wind turbines based on their states and wind parameters, maximizing energy production while respecting constraints on rotation angles and fatigue, using a simulator to calculate wake effects and energy production.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If individual turbine optimization is used, then each turbine can be controlled independently, but interactions between turbines are not considered leading to reduced energy production
Solution Approach 1:
The patent combines individual turbine control with farm-level coordination by implementing a centralized control system that optimizes the entire wind farm as an integrated system. The controller considers wake effects and interactions between turbines, merging individual turbine operations into a coordinated farm-wide strategy that maximizes total energy production.
2Productivity
If yaw control is used to steer wake effects away from downstream turbines, then energy production is improved, but dynamic mechanic loads increase
Solution Approach 1:
The patent implements dynamic control strategies that adjust yaw angles based on real-time wind conditions, turbine positions, and wake effects. The system dynamically optimizes the balance between energy production and mechanical loads by continuously adapting control parameters rather than using fixed yaw angles, thereby reducing excessive dynamic loads while maintaining improved energy production.
3Device complexity
If model-based reinforcement learning algorithms are used, then control complexity is reduced, but they assume constant wind directions and perform poorly in varying wind conditions
Solution Approach 1:
The patent implements a reinforcement learning-based control system that uses feedback from real-time wind measurements and turbine performance data to continuously optimize control decisions. The agent learns from environmental feedback and adapts its control strategy to varying wind conditions, combining the simplicity of model-based approaches with the adaptability of model-free reinforcement learning.
4Productivity
If the number of turbines is increased to expand wind farm capacity, then energy production potential increases, but physical interactions and control complexity increase
Solution Approach 1:
The patent segments the wind farm control into hierarchical levels: individual turbine control, wake effect modeling, and farm-wide optimization. This segmentation allows the system to manage large numbers of turbines by breaking down the complex control problem into manageable components, where each turbine is controlled based on its local conditions and wake effects from upstream turbines, reducing overall control complexity while maintaining high productivity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method enables precise control of large wind farms, optimizing energy production, reducing operational expenditures, and minimizing maintenance costs by accounting for interactions between turbines and dynamic wind conditions.
Implementation Method 1
Wind turbines are subject to numerous physical phenomena inside a farm: an important one is the wake effect. As wind flows through a wind turbine, wind speed decreases and turbulence increases.
Implementation Method 2
As wind flows through a wind turbine, wind speed decreases and turbulence increases. This process is called 'wake effect' and damages the farm.
Data Source
AI summary
The present invention concerns a method for controlling a wind farm comprising wind turbines, each turbine being suitable for taking a plurality of states, the method comprising the following steps:obtaining configuration data,obtaining experience data,training a model for determining actions enabling to control each turbine of the wind farm depending on the state of each turbine, the model being trained in a training environment on the basis of the experience data so as to maximize a reward function, the training environment being a multi-agent reinforcement learning environment, each agent corresponding to a different turbine of the wind farm, the reward function being relative to the energy produced by the wind farm.