NN-based NMPC Control law approximation for Maximal Wind Energy Extraction in wind turbines
ID：9
Submission ID：9 View Protection：ATTENDEE
Updated Time：2025-11-03 11:16:36
Hits：60
Poster Presentation
Start Time：2025-11-09 09:01 （Asia/Shanghai）
Duration：1min
Session：[P] Poster presentation » [P1] 1.Renewable energy system
No files
Abstract
To reduce the computational burden of online Nonlinear Model Predictive Control (NMPC) for Maximum Wind Energy Extraction (MWEE) in wind turbines, this paper proposes a neural network-based approximation method for NMPC control laws. The study investigates the principles of MPC for maximum wind energy capture, designs the prediction model's input/output parameters and structure, analyzes the impact of training data characteristics (average wind speed, turbulence intensity(TI), prediction horizon) on approximation effectiveness, and provides data-driven support for optimal model selection through comparative analysis of different network architectures. Results demonstrate that turbulence intensity and average wind speed data significantly impact model performance, while prediction horizon data exhibits minimal influence. Training with mixed turbulence level data substantially enhances model generalization capability (achieving up to 19.5% RMSE reduction), whereas mixed prediction horizon data yields limited improvement (9.6% reduction). This study proposes a neural network-based MWEE-MPC control method, providing a novel approach for real-time efficient wind turbine control under complex wind conditions.
Keywords
Maximum Wind Energy Extraction,Nonlinear Model Predictive Control,Neural Network Approximation,Control Law Approximation
Speaker
Comment submit