Enhancing Wind Power Conversion System Control Under Wind Constraints Using Single Hidden Layer Neural Network

Abstract

In the realm of wind power generation, cascaded doubly fed induction generators (CDFIG) play a pivotal role. However, the classical proportional integral derivative (PID) controllers used within such systems often struggle with instability and inaccuracies arising from wind variability. This study proposes an enhancement to overcome these limitations by incorporating a single hidden layer neural network (SHLNN) into the wind power conversion systems (WPCS). The SHLNN aims to complement the PID controller by addressing its shortcomings in handling nonlinearities and uncertainties. This integration exploits the adaptive nature and low computational demand of SHLNNs, utilizing historical wind speed and power data to form a more resilient control strategy. Through Matlab/Simulink simulations, this approach is rigorously compared against traditional PID control methods. The results demonstrate a marked improvement in performance, highlighting the SHLNN's capacity to contend with the intrinsic variabilities of wind patterns. This contribution is significant as it offers a sophisticated yet computationally efficient solution to enhance CDFIG-based WPCS, ensuring more stable and accurate energy production.