Wind Turbine Predictive Control for Real-Time Aerodynamic State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wind turbine controls rely on simplified models that limit optimal operation and lead to suboptimal performance due to inaccurate visibility into the wind turbine's state and structural dynamics, causing unnecessary wear and complex interactions that result in conservative and inefficient operation.
Innovation Solution
The use of advanced aeroelastic models and machine learning algorithms in real-time estimation and control systems to determine the current aerodynamic state and predict future performance, enabling precise actuator control through a predictive control module.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simplified models are used for real-time estimation and control, then computational complexity is reduced and control platform requirements are lowered, but measurement precision and control accuracy deteriorate
Solution Approach 1:
The control system is divided into multiple independent control loops, each handling specific functions (pitch control, torque control, load mitigation). This segmentation allows complex control tasks to be distributed across multiple simpler modules, reducing the computational burden on any single platform while maintaining overall control accuracy through coordinated operation of the loops
Solution Approach 2:
The control system dynamically adjusts the complexity and activation of different control loops based on operating conditions. During normal operation, simpler control loops are sufficient, but during abnormal conditions or when higher precision is needed, additional complex loops can be activated. This dynamic adaptation allows the system to maintain high measurement precision only when necessary, reducing average computational complexity
2Reliability
If multiple independent protection loops are implemented, then turbine protection coverage is improved, but device complexity and control interactions increase
Solution Approach 1:
Multiple protection functions are merged into a unified control architecture where loops share common estimation models and control infrastructure. This merging reduces redundant computations and simplifies the overall system while maintaining comprehensive protection coverage. The unified architecture allows loops to coordinate their actions, preventing conflicting control signals that would otherwise increase complexity
Solution Approach 2:
The control loops are designed with universal components that serve multiple functions. For example, the pitch control loop not only regulates rotor speed but also provides load mitigation and fault protection. The generator torque control similarly handles power regulation and multiple protection scenarios. This multi-functionality reduces the total number of independent loops needed while maintaining comprehensive turbine protection
3Productivity
If simplified turbine models are used for control estimation, then computational limitations are addressed, but control precision and optimal operation are compromised
Solution Approach 1:
The system pre-calculates and stores lookup tables containing pre-computed control parameters and estimation results for various operating conditions. During real-time operation, the controller quickly queries these pre-computed values based on current measurements, avoiding the need for complex real-time calculations. This preliminary action enables high-speed control responsiveness while maintaining precision by using accurate pre-computed models
Solution Approach 2:
The patent replaces complex mechanical/computational estimation processes with data-driven approaches including machine learning models and neural networks. These computational models are trained offline to capture complex turbine dynamics, then deployed for real-time estimation. This substitution allows the system to achieve high measurement precision without requiring complex real-time computational resources, as the heavy lifting is done during the training phase
Data Source
AI summary
Systems and methods are provided for the control of a wind turbine. Accordingly, a wind classification module of a controller determines a current aerodynamic state of the wind resource based, at least in part, on a current operational data set of the wind turbine. The current operational data set is indicative of a current operation of the wind turbine. A configuration intelligence module of the controller then generates an estimated configuration for a turbine estimator module and a predictive control configuration for a predictive control module based, at least in part, on the current aerodynamic state. An operation of the wind turbine is emulated via the turbine estimator module to generate a control initial state for the predictive control module. The predictive control module then determines a predicted performance of the wind turbine over a predictive interval based on the control initial state and the predictive control configuration. The predictive control module generates a set point for at least one actuator of the wind turbine based on the predicted performance, and an operating state of the wind turbine is affected via the at least one actuator in accordance with the setpoint.