Closed-Loop Wind Plant Control for Wake Loss Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wind plant control systems, particularly those using static lookup tables and open-loop models, fail to accurately predict and mitigate wake losses in wind turbines, leading to reduced energy output due to the limitations of real-world data integration and noise handling.
Innovation Solution
A model calibration and validation pipeline that translates real-world SCADA data into suitable inputs for wind plant models, using a combination of analytical calibrations and machine learning to continuously adapt and improve predictions of turbine power outputs and wake effects, enabling closed-loop control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If real-world SCADA data is used for wind plant control, then data availability and practical applicability are improved, but measurement precision and reliability deteriorate due to noise and biases in sensor data
Solution Approach 1:
The patent introduces an intermediary data processing layer that includes noise filtering, bias correction, and validation mechanisms between the SCADA sensors and the control model. This intermediary layer cleans and validates the raw sensor data before it is used for wake loss predictions, thereby maintaining practical data availability while improving measurement precision and reliability.
2Productivity
If static lookup tables are used for wake loss estimation, then computational efficiency is improved, but prediction accuracy deteriorates due to limited variables and inability to adapt to changing conditions
Solution Approach 1:
The patent transitions from static lookup tables to a dynamic predictive model that continuously adapts to changing wind plant conditions. The system uses real-time SCADA data to update wake loss predictions dynamically, allowing the model to respond to varying wind speeds, directions, and turbine configurations while maintaining computational efficiency through optimized algorithms.
Solution Approach 2:
The patent implements a feedback mechanism where actual turbine performance data is continuously compared with model predictions. The discrepancies are used to refine and update the predictive model, improving accuracy over time. This closed-loop feedback system enables the model to adapt to changing conditions and learn from operational experience while maintaining computational efficiency.
3Ease of operation
If open-loop control models are used, then system simplicity and ease of operation are improved, but control effectiveness deteriorates due to fixed assumptions about wake interactions
Solution Approach 1:
The patent implements feedback mechanisms where actual turbine performance data is continuously fed back into the control model. This allows the system to update wake interaction assumptions based on real-world observations, improving control effectiveness while maintaining relative system simplicity through automated feedback loops.
Solution Approach 2:
The control model is designed to self-update and self-correct using operational data from the wind plant. The system automatically refines its wake interaction predictions based on actual turbine performance, reducing the need for manual recalibration and maintaining ease of operation while improving reliability through continuous self-improvement.
Data Source
AI summary
Systems and methods of predicting performance of one or more wind turbines are provided. Exemplary methods include entering data inputs and analyzing and estimating the data inputs. The data inputs include nacelle position information and/or wind condition information. A wake model is determined based on the analysis and estimating of the data inputs. Wind turbine behavior predictions are generated including predicted power outputs of the wind turbines and predicted effects of waking on the predicted power outputs. The data inputs can be adjusted to improve the wind turbine behavior predictions, and the wake model can be corrected by machine learning. By focusing on an observable quantity—power—based on other observable quantities like wind speed, yaw error, nacelle position, disclosed embodiments enable optimization to start immediately after the hardware and software are installed, and as the controller operates more in the field, additional training and validation data are collected.