Wind Turbine Controller Selection via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wind turbine control methods do not effectively optimize energy production while minimizing fatigue across all wind conditions, as they fail to consider the reduction of overall fatigue as a primary objective function.
Innovation Solution
A method involving automatic online selection of controllers based on a database of simulated controllers and machine learning to determine the optimal controller for minimizing wind turbine fatigue, switching between controllers as wind conditions change, using proportional integral controllers, H∞ regulators, linear quadratic regulators, or predictive controls with different weightings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single fixed controller is used for wind turbine control, then the control system is simple and reliable, but it cannot optimize energy production across all wind conditions and fails to minimize fatigue effectively
Solution Approach 1:
The control system transitions from a static single controller to a dynamic multi-controller architecture where multiple controllers are selected in real-time based on current wind conditions. The system adapts its control strategy by choosing the most appropriate controller from the plurality of available controllers, enabling optimization across varying operational conditions while managing complexity through automated selection logic
Solution Approach 2:
The system changes the control parameter by selecting different controllers based on wind condition parameters. Each controller is optimized for specific wind condition ranges, and the system dynamically adjusts which controller is active by monitoring wind speed and other environmental parameters, thereby optimizing energy production without requiring a single complex controller to handle all scenarios
2Reliability
If complex fatigue models and multiple controllers are used, then fatigue minimization is improved, but the computational burden and system complexity increase
Solution Approach 1:
The system performs preliminary classification of wind conditions and pre-selects appropriate controllers from the plurality available. By categorizing wind conditions in advance and having pre-configured controllers for each category, the system reduces real-time computational burden while maintaining effective fatigue minimization strategies for each operational regime
Solution Approach 2:
The invention introduces an intermediary controller selection mechanism that bridges the gap between complex fatigue models and actual control execution. This intermediary layer analyzes wind conditions, selects the most appropriate controller from multiple candidates, and manages the switching between controllers, thereby simplifying the overall system architecture while preserving the benefits of complex fatigue-aware control strategies
Data Source
Figure 1~3
AI summary
The present invention relates to a method for controlling a quantity of a wind turbine through the online automatic selection of a controller which minimizes wind turbine fatigue. To this end, the method is based on a database (BDD) (constructed offline) of simulations of a list (LIST) of controllers, and on an online machine learning step in order to determine the optimal controller in terms of wind turbine (EOL) fatigue. Thus, the method allows the online automatic selection of controllers on the basis of a fatigue criterion, and switching between the controllers according to changes in the measured wind conditions.