Wind Park Optimization via Data-Driven Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wind park optimization methods are static, reliant on physics models and simulations, and fail to adapt to real-time conditions, leading to suboptimal performance due to wake effects and sensor errors.
Innovation Solution
A data-driven method for dynamic real-time optimization of wind park performance, which maps wind condition data and operational parameters to a target parameter, allowing for self-learning optimization without relying on simulations or physics models, and compensates for sensor errors by updating operational parameters based on measured posterior conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If physics models and simulations are used for wind park optimization, then optimization accuracy can be improved, but computation time and complexity increase significantly
Solution Approach 1:
The patent replaces complex physics models and simulations with a data-driven machine learning approach. Instead of using computational fluid dynamics or physics-based wake models that require extensive computation, the system uses historical sensor data from the wind park to train machine learning models that predict wake effects and optimize turbine operations in real-time, dramatically reducing computation time while maintaining accuracy
Solution Approach 2:
The patent creates simplified digital representations of the wind park's operational characteristics by training machine learning models on historical data. These models capture the essential wake dynamics and turbine responses without requiring full physics simulations, enabling fast real-time optimization while preserving the key behavioral patterns of the system
2Device complexity
If static optimization methods are used, then device complexity is reduced, but adaptability to real-time conditions deteriorates
Solution Approach 1:
The patent implements dynamic optimization by continuously updating machine learning models with real-time sensor data from the wind park. The system adapts to changing wind conditions, turbine performance variations, and environmental factors by retraining models with current data, enabling the optimization strategy to respond dynamically to real-time conditions while maintaining manageable complexity through automated data processing
Solution Approach 2:
The patent establishes a closed-loop feedback system where sensor measurements from the wind park are continuously fed into machine learning models that predict optimal turbine operations. The system monitors actual performance, compares it with predictions, and uses the discrepancies to refine and update the models, creating a self-improving adaptive optimization system that maintains low complexity through automated feedback processing
3Measurement precision
If physics models are relied upon, then theoretical accuracy is improved, but compensation for sensor errors and systematic deviations becomes difficult
Solution Approach 1:
The patent implements self-service error compensation by training machine learning models directly on historical sensor data from the wind park. The models learn the systematic deviations and measurement errors inherent in the specific sensor suite used, automatically compensating for these errors through data-driven calibration. This self-learning approach allows the system to adapt to sensor characteristics without requiring external calibration or complex error correction models
Solution Approach 2:
The patent transforms the optimization approach from physics-model-based with fixed parameters to data-driven with adaptive parameters. By training machine learning models on historical data, the system learns optimal parameter values that inherently account for sensor errors and systematic deviations. The models adjust their internal parameters based on observed patterns in the data, enabling automatic compensation for measurement inaccuracies while maintaining theoretical accuracy
Data Source
Figure 1a
Figure 1b
Figure 2
AI summary
A method for dynamic real-time optimization of the performance, in particular of the power output, of a wind park (100), wherein the wind park (100) comprises a plurality of wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g), the method comprising the steps of: a) providing, for a plurality of subsets (12a, 12b, 12c, 12d, 12e) of the wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g) of the wind park (100), a mapping of a data vector to a target parameter, said target parameter being indicative of the performance of the subsets (12a, 12b, 12c, 12d, 12e) of the wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g), said data vector comprising wind condition data and operational parameters for the wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g) of the subsets (12a, 12b, 12c, 12d, 12e) of wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g), b) selecting a subset (12a, 12b, 12c, 12d, 12e) of the plurality of subsets (12a, 12b, 12c, 12d, 12e) of wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g) for optimization, c) determining the wind condition data for the selected subset (12a, 12b, 12c, 12d, 12e) of wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g), d) determining the operational parameters for the wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g) of the selected subset (12a, 12b, 12c, 12d, 12e) by at least one of - selecting operational parameters which maximize or minimize the value of the target parameter for the determined wind condition data, and/or - selecting operational parameters for which the value of the target parameter for the determined wind condition data is unknown, the method further comprising the steps of e) applying the determined operational parameters to at least one of the wind turbines (10, 10a, 10b, 10c, 10d, 10e, 10f, 10g) of the selected subset (12a, 12b, 12c, 12d, 12e), f) after a first time interval, measuring posterior wind conditions and determining a posterior value of the target parameter for the selected subset (12a, 12b, 12c, 12d, 12e), g) updating the mapping with at least one of the posterior wind conditions and/or the posterior value of the target parameter of step f) for the applied determined operational parameters, and h) repeating, after a second time interval, steps b) to h).