Per Unit-Based Neural Network Control for MPPT of Wind Turbines: Reducing Data Needs and Enhancing Generalization

Abstract

Artificial Neural Networks (ANNs) offer a highly adaptive and precise approach to tracking the Maximum Power Point (MPPT) of wind turbines. Despite their advantages, achieving effective generalization with ANNs often requires significant computational resources—especially when dealing with large datasets—and depends critically on the careful selection of input and output variables to ensure optimal performance.
In this study, we developed an ANN-based MPPT controller for wind energy conversion systems. The proposed network uses wind speed, mechanical power, and turbine rotational speed as input variables, while the output is the rotational speed expressed in Per Unit (PU). This PU-based representation not only simplifies the training process but also enhances the learning efficiency and generalization capability of the network across different wind turbine configurations.
Simulation results demonstrate that the proposed ANN controller provides accurate and robust performance in tracking the maximum power point, highlighting its potential as an effective solution for intelligent wind energy management.