Enhancing Model Predictive Control of a 3 MW Wind Turbine with Machine Learning

Countries are heavily investing in renewable energy to meet climate targets, with wind energy playing a crucial role, especially in the EU where it accounted for 37.5% in 2022. The industry is focusing on building larger and more numerous wind turbines (WTs) to boost green energy production. This expansion increases the complexity of WT operations, requiring solutions to reduce mechanical load alternation, extend service life, and maintain acceptable sound emissions while ensuring power output. Model predictive wind turbine control (MPWTC) is a promising approach that integrates competing control objectives and constraints into a single real-time optimization problem using a predictive process model of WT dynamics.

Incorporating machine learning into MPWTC enhances the prediction of complex WT behaviors, such as mechanical thrust load changes. See how to use MATLAB^® and Simulink^® to integrate a new control objective into MPWTC, starting with extending the WT process model using a local linear neuro-fuzzy model trained on high-fidelity simulation data. This enhanced model is then applied in MPWTC to predict thrust changes in real-time. The approach is validated through Simulink simulations and real-world experiments using a full-scale 3 MW WT, showing its practical viability. Discover how to use automated code generation with Simulink Coder™ for a Bachmann MH230 PLC to implement the control on the WT. Finally, explore future trends in wind turbine control and the potential advancements enabled by AI.